Service outages: impact, recovery & what to do
Understand a reported outage before changing your setup. Selected official reports, explained with the affected features, regions and safe next steps.
Editorial review ·
Review target: every 3 hours. This is a selected editorial bulletin, not an exhaustive or continuous live feed.
Reports reflect the checks below. Conditions may change before the next review.
Review overdue — treat the entries as last-known reports and check the official sources for the current situation.
Times use your device’s time zone when JavaScript is available; UTC is also shown. Date-only source stamps have no inferred time.No matching reports. This does not mean there is no outage. Try a broader search or open the live monitor.
Needs recheck
US meeting disconnects: recovery under monitoring
- Who / what is affected
- US region · a subset of Zoom Meetings users
Zoom reported intermittent meeting disconnects at 19:16 UTC on October 22 and identified the cause at 19:32. At 20:05 it said the degradation had been resolved, but the incident's formal status remains Monitoring. The notice concerns a subset of users in the US region, not every meeting or every country, and does not establish whether your recording or meeting data was saved.
Safe next step
Preserve the meeting ID, error and time, and confirm whether the meeting or recording is still active before creating duplicates. Follow your organization's approved fallback if needed. Do not disable security controls, delete the app's data or change unrelated DNS settings because of this regional notice.
Needs recheck
Real-time connections and messaging: reported recovered
- Who / what is affected
- Some Slack customers · connections, message delivery and history loading
Slack's retrospective notice reports disrupted connections, delayed messages and conversation-history loading errors from 16:08 UTC on October 22, with normal service restored by 16:26. Many simultaneous reconnections briefly strained messaging services. The notice was published at 18:50 UTC; that is not the outage start. Slack plans a full investigation, so the technical trigger and preventive actions are not yet a final root-cause report. No country list is supplied.
Safe next step
Check whether a message was actually delivered before sending it again, and keep a copy of unsent work. Record the workspace, error and time if problems persist. Do not delete conversation history, reset the workspace or disable account protections to address a provider-side connection incident.
Needs recheck
Dots message failures: provider reports recovery
- Who / what is affected
- ChatGPT Dots · impact primarily on Enterprise customers
OpenAI reported failures sending messages to Dots at 18:51 UTC on October 22. Its 18:59 update said the majority of impact affected Enterprise customers and mitigation was being implemented. At 20:55 it formally marked the incident Resolved and reported full recovery of impacted services. This is not a report that every ChatGPT feature, subscription or API failed, nor confirmation of an individual message's outcome.
Safe next step
Check the conversation before resending and preserve unsent work and error details. Use only an organization-approved alternative for confidential data. Do not share credentials, delete workspaces or weaken account protections to work around this reported service issue.
Needs recheck
New ad account admin assignment: mitigation under monitoring
- Who / what is affected
- New OpenAI ad accounts · Ads Manager and Ads API
OpenAI identified failures assigning a new ad account's creator as administrator at 19:36 UTC on October 22. After a 21:30 update saying mitigation was being prepared, it reported applying the mitigation at 23:45 and moved the incident to Monitoring, not Resolved. The official page lists Ads Manager and Ads API. This does not confirm that every affected account now has the correct role, establish loss of existing administrators or indicate a general ChatGPT outage.
Safe next step
Keep the account ID, creation time and error, and ask official support to verify the assignment. Avoid creating duplicate ad accounts or repeating billing actions to obtain access. Never share login details, use someone else's administrator account or disable two-step verification as a workaround.
Needs recheck
Log drains: rollback followed by reported resolution
- Who / what is affected
- Some regions · routed logs for a subset of components
Supabase reported at 18:03 UTC on October 22 that log-drain customers in some regions might be missing routed logs for a subset of components. At 18:22 it said a rollback restored normal operation; at 18:50 it formally resolved the incident. Regions and affected components are not fully enumerated. This is distinct from the API Gateway and Management API incidents and does not confirm that every missing historical log was recovered.
Safe next step
Keep independent application logs, request IDs and time ranges, and check the destination before rerunning an operation. Ask official support about any remaining gaps. Do not repeat production writes just to generate logs, delete projects or weaken logging and security policies.
Needs recheck
API Gateway demand spike: further mitigations planned
- Who / what is affected
- Multiple regions · Data APIs and Project Creation
Supabase linked API Gateway request issues to demand on its internal backplane and said the initial mitigations managed the spike at 17:57 UTC on October 22. At 18:17 it announced additional mitigations to reinforce the backplane and said the incident would remain Monitoring until all were deployed. Their deployment is not yet confirmed. The page lists Data APIs and Project Creation in 18 regions; this is distinct from the lifecycle and Management API incidents, not proof that every running database or request failed.
Safe next step
Check your project region, request ID and actual request or job outcome before a limited retry. Avoid duplicate project creation, repeated restarts or an unplanned production restore; preserve errors and timestamps for official support.
Needs recheck
GitHub Projects processing improves; new items may still be delayed
- Who / what is affected
- GitHub Projects · new items and project updates · other affected services recovered
GitHub's October 22 incident is now titled Project updates delayed. Its 22:07 UTC update described recovery and work to add capacity; at 23:48 it said update processing had improved but newly added items could still take longer to appear. The formal status remains Investigating despite the recovery wording. The other affected services had recovered earlier. This is the same incident previously covering Issues, Pull Requests and search, not a new outage or proof of completed Projects recovery, and is separate from the resolved Actions-results incident.
Safe next step
Check whether the issue, pull request, project edit or downstream action was saved before repeating it. Keep the URL, error and time for support. Do not bypass required checks, force-push, delete repositories or rotate credentials merely to fix a delayed page or project update.
Needs recheck
Western Europe Durable Objects: fix under monitoring
- Who / what is affected
- Western Europe · Durable Objects and related services
Cloudflare reported elevated error rates for Durable Objects and related services in Western Europe at 17:51 UTC on October 22. At 18:19 it said a fix had been implemented and moved the incident to Monitoring, not Resolved. The notice does not enumerate all affected countries, establish a worldwide Cloudflare outage or confirm data loss. It remains distinct from the earlier Lisbon and North American incidents.
Safe next step
Match the failing request to the affected service and region. Preserve request IDs and errors, and check whether a state-changing request completed before retrying. Do not delete stored state, disable protections or change unrelated DNS settings as a workaround.
Needs recheck
WARP audit-log delivery: fix implemented, monitoring continues
- Who / what is affected
- WARP API audit logs · Zero Trust · some dependent services
Cloudflare reported delayed or missing audit-log delivery for some WARP API customers and possible effects on dependent services. It identified capacity pressure in a shared internal data pipeline in PDX. At 17:50 UTC on October 22, it said a fix was implemented and moved to Monitoring, not Resolved. Missing audit records do not prove that no activity occurred, that all WARP connections failed or that a security breach took place.
Safe next step
Keep WARP and Zero Trust protections enabled. Preserve independent logs and request times, and account for gaps when reviewing activity. Ask your administrator or official support to assess missing records; do not bypass security policies simply to obtain logs.
Needs recheck
Network Error Logging: provider reports resolution
- Who / what is affected
- NEL processing and analytics · not website traffic
After partial service restoration at 20:08 UTC on October 22, Cloudflare reported a fix under monitoring at 21:36 and formally resolved the NEL incident at 22:00. The original notice said affected reports would not appear in NEL analytics and explicitly excluded website traffic and other services. Resolution does not confirm that missing historical analytics were backfilled. The original undated start-time wording remains ambiguous and is not treated as a verified dated onset.
Safe next step
Retain independent browser, application and server diagnostics. Do not interpret an empty NEL chart as proof that requests had no errors, and do not change DNS or disable security protections to restore analytics.
Needs recheck
Functions and Edge Functions log delays: reported resolved
- Who / what is affected
- Logs from 13:58–15:35 UTC on October 22 · UI and log drains
Netlify formally resolved this incident at 20:10 UTC on October 22. Its earlier notice concerned delayed Functions and Edge Functions logs from 13:58–15:35 UTC appearing in the UI and log drains while a backlog was processed. Function and Edge Function requests were explicitly unaffected. This is separate from the earlier build-log incident; provider resolution alone does not verify that every individual log is present at your destination.
Safe next step
Preserve invocation IDs, timestamps and independent evidence of outcomes. Do not rerun production functions merely to make logs appear; check whether an operation already completed before retrying and use official support if records remain missing.
Needs recheck
Actions results incident reported resolved
- Who / what is affected
- GitHub Actions · workflow and step results · repository updates
GitHub moved this October 22 Actions-results incident to Monitoring at 15:17 UTC, then reported workflow and step result processing and repository updates recovered and formally marked it Resolved at 15:18 UTC. A detailed root-cause analysis is still pending. This does not verify every individual workflow or deployment outcome, and it is separate from both the October 19 retrospective and the later October 22 Issues/search investigation.
Safe next step
Check the run ID, commit, step state and actual deployment or downstream outcome before rerunning a workflow. Preserve errors and times, and avoid duplicate production changes. Do not bypass required checks, force-push code, delete repositories or rotate credentials solely because a result is delayed.
Needs recheck
Management API and supabase.com login incident reported resolved
- Who / what is affected
- Management API · ap-southeast-1 / eu-west-1 / us-east-1 / us-west-1 · Dashboard
Supabase formally resolved this incident at 13:42 UTC on October 22 after error rates and latency remained normal following mitigation. Its 12:07 UTC notice had reported mitigation and normal rates while monitoring stability. Earlier reports concerned Management API errors, dependent operations such as project creation, and supabase.com login timeouts, with four API regions and Dashboard listed. Resolution does not verify each management job, and this was not confirmation of an outage of every running database or customer application's end-user authentication. The separate EU lifecycle incident remains under monitoring; this resolution must not be applied to it.
Safe next step
Keep the project, region, operation ID, time and error details. Check whether a management job already completed before retrying. For login failures, verify the official address and status; do not disable MFA, rotate keys, delete projects or start an unplanned production restore solely to bypass this provider incident.
Needs recheck
Online Checkout incident reported resolved; Global region listed
- Who / what is affected
- Online Checkout · PP-LIVE-110330 · official affected-region label: Global
PayPal's official status RSS published a Detected Resolved notice for Online Checkout at 08:14 UTC on October 22. It says engineers resolved the issue and mitigated the impact, with affected regions listed as Global. The feed does not give an exact outage start or recovery time; its publication time is not proof of either. This does not establish that every country, payment method or transaction failed, or confirm the outcome of an individual payment or refund.
Safe next step
Check the payment ID and actual status in the official PayPal account or approved integration before trying again or asking a customer to pay again. Reconcile completed, pending and failed transactions separately. Preserve receipts and errors; avoid duplicate charges, refunds or orders based only on a timed-out checkout screen.
Needs recheck
Connector-error incident reported resolved
- Who / what is affected
- claude.ai / Claude Cowork · connectors · affected countries not specified
Claude reported elevated connector errors at 03:50 UTC on October 22, deployed a fix under monitoring at 07:28 UTC and explicitly marked the claude.ai / Claude Cowork incident resolved at 08:07 UTC. This supersedes the earlier investigation notice. It does not prove completion of every connector action or establish that all APIs or Claude Code sessions were unavailable. A common cause with Notion's separate MCP incident is not confirmed.
Safe next step
Keep the connector name, task ID, time and error details. Check the connected service and whether an action completed before retrying. Do not delete working connections, rotate credentials or broaden permissions solely because of this provider notice; preserve unsaved work and follow your organization's access policy.
Needs recheck
EU project lifecycle incident reported resolved
- Who / what is affected
- EU region in the notice · Management API eu-west-1 component · create / restore / pause
At 15:56 UTC on October 22, Supabase formally marked this EU project lifecycle incident Resolved after error rates remained normal following mitigation. The earlier notice reported errors since 23:00 UTC on October 21, with the eu-west-1 Management API component listed. Provider recovery does not verify the outcome of every affected creation, restore or pause job. This remains separate from the October 21 all-regions incident, the earlier multi-region Management API report and the new API gateway demand incident; do not infer a common cause or an outage of all running databases.
Safe next step
Check the project ID, region and lifecycle job state before retrying creation, pausing or restoration. Keep errors and timestamps for official support. Do not delete projects, issue repeated restarts or start an unplanned production restore as a workaround; follow an assessed recovery plan.
Needs recheck
Notion MCP query_data_source in Claude clients reported available again
- Who / what is affected
- Notion MCP · query_data_source tool · Claude clients
Notion reported the MCP query_data_source tool temporarily unavailable in Claude clients at 03:23 UTC on October 22 and explicitly marked the issue resolved at 04:42 UTC, saying the tool is available again. This does not establish an outage of all Notion, all MCP tools or every Claude connector, or confirm completion of each interrupted query. It is separate from the earlier infrastructure migration and Claude's connector-error report; a common cause is not confirmed.
Safe next step
Keep the query, tool name, time and error details. Check the result and any downstream task before retrying. Do not recreate connections, rotate integration tokens or broaden workspace permissions solely because of this incident; use only access already authorized for the data source.
Needs recheck
ChatGPT conversation and Responses API incident reported resolved
- Who / what is affected
- Conversations / Responses API · India, Japan and Korea named in revised updates
OpenAI explicitly reported full recovery and marked this incident resolved at 08:50 UTC on October 22, following investigation at 02:18 UTC and mitigation under monitoring at 02:42 UTC. The revised title refers to ChatGPT conversations and Responses API errors; revised earlier updates name users in India, Japan and Korea. This replaces the earlier conversion-failures wording, not a newly separate incident. It does not confirm a worldwide outage or completion of each interrupted conversation or API operation.
Safe next step
Preserve prompts, partial outputs, error responses and request IDs. Check existing conversations and completed operations before retrying a write or billable API request; repeat only failed work. Do not delete conversations, rotate API keys or move restricted data to an unapproved service solely because of this provider incident.
Needs recheck
Checkout Pro performance incident reported resolved
- Who / what is affected
- Online payments · Checkout Pro · countries not specified in the notice
Mercado Pago reported degraded Checkout Pro performance at 01:59 UTC on October 22 and formally marked the incident Resolved at 02:19 UTC. The notice does not name affected countries or establish failures in every payment method. This is separate from the Argentina in-store QR incident and does not prove that each payment, refund or notification completed.
Safe next step
Check the payment ID and its actual status in the account or approved integration before asking a customer to pay again. Reconcile completed, pending and failed transactions separately, and keep the error and time. Avoid repeated charges, duplicate refunds or recreating orders merely because the checkout page previously failed or timed out.
Needs recheck
Argentina: new in-store QR payment incident reported resolved
- Who / what is affected
- Argentina (MLA) · in-store QR payments
Mercado Pago reported increased declines for in-store QR payments in Argentina at 01:52 UTC on October 22 and marked the incident Resolved at 02:16 UTC, corresponding to 22:52–23:16 on October 21 in Argentina. This is a separate report from the October 19 QR and October 22 UTC Checkout Pro incidents, not proof of a continuous outage or failures in all payment methods or countries.
Safe next step
Confirm the transaction's actual status with its payment ID before trying again or switching payment methods. A declined or timed-out screen alone does not prove that no charge occurred. Preserve receipts and error details, reconcile pending payments and avoid repeated charges or duplicate refunds.
Needs recheck
Portland and Seattle congestion incident reported resolved
- Who / what is affected
- United States · Portland (PDX) / Seattle (SEA) facilities
Cloudflare investigated network congestion in Portland and Seattle at 00:23 UTC on October 22, reported a fix under monitoring at 01:08 UTC and formally marked the incident Resolved at 02:09 UTC. The notice names these facilities and regional impact, not an outage of every connection in the northwestern US or Cloudflare worldwide. It does not confirm completion of each interrupted application request.
Safe next step
Compare the URL, error, region and time with the provider notice and your own logs. Before repeating a state-changing request, check whether it already completed. Do not reinstall applications, disable security controls or change DNS solely because of this regional provider incident.
Needs recheck
Lisbon: Durable Objects incident reported resolved
- Who / what is affected
- Lisbon (LIS), Portugal · Durable Objects handled by this facility
Cloudflare formally marked this Lisbon Durable Objects incident Resolved at 00:35 UTC on October 22, after the investigation notice at 23:50 UTC on October 21. The resolution concerns this specific incident, not proof that every Lisbon facility service or interrupted operation is healthy or completed. The notice does not establish an outage across Portugal or all Durable Objects locations and remains separate from Eastern North America.
Safe next step
Record the request ID, error, location and time, and compare them with application logs and the current official notice. Check whether a state-changing operation completed before retrying, using normal duplicate-prevention safeguards. Do not reset object state, delete storage or change DNS merely to bypass this regional provider issue.
Needs recheck
Claude request-error incident reported resolved
- Who / what is affected
- Notice names Claude Fable 5 · lists claude.ai, API, Claude Code and Cowork
Claude investigated elevated errors for Claude Fable 5 at 19:49 UTC on October 21 and marked the incident Resolved at 20:29 UTC. The notice lists claude.ai, the Claude API, Claude Code and Claude Cowork. Its closing text refers to other Claude models, without naming them; this does not establish that every model, request or feature failed. Provider recovery does not verify each interrupted operation.
Safe next step
Preserve prompts, local changes, partial outputs and request IDs. In an agent workflow, check file changes, command results and completed actions before resuming so that writes or deployments are not duplicated. Do not delete project data or move restricted data to an unapproved service merely to bypass a provider error.
Needs recheck
Eastern North America: R2 and Durable Objects errors reported resolved
- Who / what is affected
- Eastern North America (ENAM) · some R2 / Durable Objects customers
Cloudflare revised the October 21 error window for some R2 and Durable Objects customers in Eastern North America to approximately 19:02–19:05 UTC, replacing its earlier 19:19–19:28 report. The revised timeline shows a resolved entry at 19:05 UTC and a later resolution notice and last update at 21:21 UTC. The incident remains Resolved; these times are not evidence of a continuous outage between the entries or completion of every failed write.
Safe next step
Match the region, request ID and time with application logs. Check whether an object write or state-changing operation completed before retrying it, using the application's normal duplicate-prevention safeguards. Do not delete storage, reset object state or change DNS solely because of this regional incident.
Needs recheck
Bitbucket availability incident reported resolved
- Who / what is affected
- Bitbucket · Git over SSH/HTTPS · website/API · Pipelines and related services
Bitbucket investigated a disruption at 17:22 UTC on October 21, reported recovery under monitoring at 17:51 UTC, and formally marked the incident Resolved at 18:01 UTC with normal service for affected customers. The notice covers several Git, web and automation components. Provider recovery does not prove that each interrupted push, pipeline or webhook delivery completed.
Safe next step
Preserve local commits and unpushed work. Check the remote branch, commit and pipeline or delivery ID before retrying. Do not delete the repository, discard changes or repeatedly trigger deployments solely because an earlier request timed out.
Needs recheck
India: increased latency incident reported resolved
- Who / what is affected
- India · some traffic · CDN/Cache
Cloudflare reported increased latency affecting some traffic in India at 16:42 UTC on October 21, implemented a fix and moved to monitoring at 18:51 UTC, and formally marked the incident Resolved at 19:57 UTC. The notice lists CDN/Cache. It does not describe all Indian connections failing or a worldwide outage, and is separate from scheduled Mumbai maintenance.
Safe next step
Record the affected URL, region, time and error or latency details. Compare them with the official notice before changing the site or network. Do not delete site data, disable security controls or change DNS simply to bypass a reported provider latency issue.
Needs recheck
gpt-5.6-luna API error incident reported resolved
- Who / what is affected
- gpt-5.6-luna API requests · Chat Completions / Responses
OpenAI reported elevated gpt-5.6-luna API errors at 15:34 UTC on October 21, moved to monitoring at 15:47 UTC and explicitly reported full recovery of impacted services at 16:44 UTC. The page lists Chat Completions and Responses. This is separate from the FedRAMP incident, not proof that every model or ChatGPT was unavailable or that each request completed.
Safe next step
Keep request IDs, error responses and partial outputs. Check whether work already completed before retrying a write or billable request, and retry only the failed operation. Do not rotate API keys or delete project data solely because of this provider incident.
Needs recheck
Project creation, pausing and restoration incident reported resolved
- Who / what is affected
- All regions in the notice · project lifecycle operations
Supabase investigated project lifecycle failures in all regions at 15:36 UTC on October 21, reported a fix and monitoring at 16:12 UTC, and explicitly confirmed normal operations and resolution at 16:51 UTC. The notice concerns operations such as creating, pausing and restoring projects, not a confirmed outage of every running database. It does not verify the result of each individual job.
Safe next step
Check the project ID and existing job state before retrying creation, pausing or restoration. Keep error details and consult official support if a job is still pending. Do not create duplicate projects or start an unplanned production restore as a workaround; follow an assessed recovery plan.
Needs recheck
Pull Requests, Actions and webhook incident reported resolved
- Who / what is affected
- Pull Requests / Actions / Webhooks · audit-log event backlog
At 19:22 UTC on October 21, GitHub reported that Pull Request updates, Actions and webhook deliveries had recovered and audit-log delays were below 40 minutes, within its healthy threshold, while the remaining backlog continued to process. It then formally marked the incident Resolved. This is not confirmation of zero audit-log delay or completion of every operation or audit event. A detailed root-cause analysis is still pending.
Safe next step
Check Pull Request, workflow and delivery IDs before retrying, and reconcile audit events after the backlog clears. Missing or delayed audit events do not prove that an action never happened. Avoid duplicate merges, deployments or redelivery. Do not disable auditing, rotate secrets or broaden repository permissions solely because events are delayed.
Needs recheck
Project-sidebar and navigation incident reported resolved
- Who / what is affected
- Jira / JSM / JPD / JWM · project sidebar and navigation
At 15:45 UTC on October 21, Atlassian formally marked the project-sidebar and navigation incident Resolved and reported normal operation for affected customers. The follow-up lists Jira, JSM, JPD and JWM. This supersedes the earlier Monitoring notice, but does not prove that each interrupted edit or operation was saved or completed.
Safe next step
Preserve unsaved issue edits and comments. If available, use a known issue or project URL that you already have permission to access, and check whether changes were saved before resubmitting them. Do not change project permissions or delete local data simply to repair a missing navigation menu.
Needs recheck
FedRAMP Astra/Sol API error incident reported resolved
- Who / what is affected
- OpenAI FedRAMP environment · Astra / Sol API errors
OpenAI explicitly reported full recovery of the affected FedRAMP Astra/Sol services and marked the incident Resolved at 19:34 UTC on October 21. This follows mitigation and Monitoring displayed at 16:46 UTC in the revised timeline. The notice concerns the FedRAMP environment, not a general ChatGPT, Codex or public API outage, and does not verify the result of each individual request.
Safe next step
Check the approved environment and model, and preserve request IDs and error details. Confirm whether an operation completed before retrying a write or billable request. Do not move regulated or restricted data to a consumer service or another endpoint simply to bypass the incident; use only a fallback approved by your organization.
Needs recheck
Build-log delay incident reported resolved
- Who / what is affected
- Build Pipeline · Netlify Application UI · log visibility
Netlify reported completed build-log backlog processing at 09:35 UTC on October 22 and expected consistent access to historical and new logs while monitoring for recurrence. Its revised timeline shows a formal Resolved notice at 10:12 UTC; that entry was created at 10:28 UTC and last edited at 10:28:54 UTC. These provider report times are distinct from the earlier estimated deadlines and do not prove completion of each build, deployment or write. The notice concerns log availability, not confirmation that all hosted sites had been offline.
Safe next step
Check the deploy status and result separately from whether its logs have appeared. Keep the deploy ID, commit and error details. Check official updates before canceling or repeatedly triggering a production build; missing logs alone are not a reason to roll back a working deployment or repeat automated actions.
Needs recheck
ChatGPT, Codex and API performance incident reported resolved
- Who / what is affected
- ChatGPT features · Codex Web / API / CLI · multiple API components
OpenAI's revised official timeline displays investigation at 02:26 UTC on October 21, mitigation and monitoring at 02:48 UTC, and full recovery of impacted services at 03:28 UTC. These displayed times differ from the earlier version of the notice. The page lists ChatGPT, Codex and multiple API features. This does not establish that every model or request was completely unavailable, or that each interrupted chat, agent run or API operation completed.
Safe next step
Preserve unsaved prompts, work and code changes. Check the conversation, task history and outputs before rerunning work, especially operations that can write data or incur charges. Retry only the failed operation after checking its state. Do not rotate API keys or delete local project data solely because of this provider incident.
Needs recheck
us-east-2 restarting-state incident reported resolved
- Who / what is affected
- us-east-2 · some projects in the restarting state
The incident affected some us-east-2 projects stuck in the restarting state. At 04:47 UTC on October 21, Supabase said the fix had fully rolled out and projects were no longer stuck. At 05:39 UTC it explicitly resolved the incident, reporting all affected projects fixed and no recurrence. Provider recovery does not prove that each interrupted client operation, write or job completed. The notice does not establish a worldwide outage or data loss.
Safe next step
Check the project ID, region and current state against the official notice. Avoid repeatedly restarting, deleting or recreating the project as a workaround. Save error details and timestamps for official support, and check whether writes or jobs already completed before retrying them. Use an assessed recovery plan rather than an unplanned production restore.
Needs recheck
Brazil: separate evening in-store QR incident reported resolved
- Who / what is affected
- Brazil (MLB) · in-store QR payment declines · October 20, 22:02–22:35 UTC
Mercado Pago reported increased in-store QR payment declines in Brazil at 22:02 UTC on October 20 and explicitly marked this incident resolved at 22:35 UTC (19:02–19:35 in Brasília). This is separate from the earlier 03:32–03:55 UTC incident that day, not proof of a continuous all-day outage. It does not establish failures in every payment method or country, or verify each transaction.
Safe next step
Check the payment ID, time and final status in the official account before retrying or asking a customer to pay again. Reconcile the merchant and customer records to avoid duplicate charges. Do not treat a service recovery notice as proof that an individual payment succeeded or failed.
Needs recheck
Precautionary certificate work; no compromise evidence reported
- Who / what is affected
- eMudhra certificate authority · preventative security notice · not an outage report
At 19:00 UTC on October 20, Notion said it was assessing potential exposure to an eMudhra certificate-authority security issue and updating affected software, certificate bundles and validation components. Notion said it had not identified evidence that it was compromised in connection with this issue. The notice is Investigating and describes preventative security work, not a service outage or proof that your account was breached.
Safe next step
Use only official Notion app updates and your organization's IT guidance. If a certificate warning appears, do not bypass it, disable certificate verification or install untrusted certificates. Preserve unsynced work and consult official support or IT if access fails; do not treat this notice alone as evidence of an account breach.
Needs recheck
Installing other members' connections: resolution reported
- Who / what is affected
- Workspace-scoped connections · installation by someone other than the creator
At 14:43 UTC on October 21, Notion explicitly marked the issue installing connections created by other workspace members resolved. The earlier problem made those connections absent from Browse; creators could still install their own. This recovery concerns that installation issue, not confirmation that every existing integration works or that the separate infrastructure migration has finished.
Safe next step
Confirm who created the connection and whether it is workspace-scoped, then check the official update or contact your workspace administrator. Do not rotate tokens, recreate integrations, remove working connections or broaden permissions solely to make an item appear; those changes can create duplicates or disrupt existing access.
Needs recheck
Verified-bot categorization incident reported resolved
- Who / what is affected
- Bot Management · limited number of verified bots · categorization, not confirmed blocking
Cloudflare reported incorrect traffic categorization involving a limited number of verified bots at 19:35 UTC on October 20 and explicitly marked the incident resolved at 20:16 UTC. The notice concerns Bot Management categorization; it does not prove that every crawler was blocked, that all automated traffic was malicious, or that your site's rules caused the issue.
Safe next step
Compare request times and existing Bot Management or security-event records. Verify the actual allow, challenge or block action before attributing an access failure to this notice. Do not disable bot protection, allow all automated traffic or change country-blocking rules solely because of this resolved report.
Needs recheck
London: recovery and revised timeline in preliminary incident report
- Who / what is affected
- London · europe-west2-b zone · regional errors in europe-west2
Google's October 22 preliminary report describes a power and backup-control failure in europe-west2-b, with knock-on regional errors and queues. The main impact window is 06:00–12:10 UTC on October 20; 06:08 was the alert time. Over 99% of Compute Engine VMs were healthy by 12:10 UTC, and the report describes recovery of the remaining isolated workloads before a closure notice by 14:03 UTC. The public record's end time now also reads 12:10 UTC, replacing 14:00 UTC; that impact-window endpoint is distinct from the closure notice. This remains a preliminary, revisable report, not a final RCA or proof that every interrupted write or job completed.
Safe next step
Match the project, region, zone and service to the notice. Check logs and whether writes or jobs already completed before retrying. If still affected, open an official support case with the workload ID, error and time. Do not move or recreate production workloads solely on this notice; use an assessed failover plan rather than disabling protections or making unrelated DNS changes.
Needs recheck
Agent Runner failures: recovery reported
- Who / what is affected
- Agent Runners · no specific countries named
Netlify reported increased Agent Runner failures at 07:47 UTC on October 20, moved to monitoring at 08:04 UTC and marked the incident resolved at 08:55 UTC. The notice concerns Agent Runners, not a confirmed outage of every Netlify-hosted site. Provider recovery does not confirm that each interrupted run completed.
Safe next step
Check the run history, logs and any generated changes before restarting a task to avoid duplicate actions. Preserve your work and keep the run ID, error and time for official support if failures continue. Do not broaden permissions or disable security controls as a workaround.
Needs recheck
Zero Trust connectivity and Access API incident resolved
- Who / what is affected
- Access API · Gateway traffic with identity policies · Dashboard
Cloudflare reported the Access API fully restored at 06:14 UTC on October 20 and marked this Zero Trust connectivity incident resolved at 06:32 UTC. It had previously confirmed recovery of Access logins and Secure Web Gateway connectivity. The incident affected Gateway traffic with identity policies and the Dashboard. This was a separate incident from the Gemini DLP enforcement issue; provider recovery does not confirm that each interrupted API operation completed.
Safe next step
Check whether the failure involves an Access API operation or a user login. Verify the current configuration before retrying API writes to avoid duplicate changes; give your IT team the error and time if problems continue. Do not remove identity policies, bypass Gateway, disable protections or make unrelated DNS changes.
Needs recheck
October 20 Brazil in-store QR payment incident resolved
- Who / what is affected
- Brazil (MLB) · in-store QR payments · separate October 20 report
Mercado Pago reported increased in-store QR payment declines in Brazil at 03:32 UTC on October 20 and marked the incident resolved at 03:55 UTC (00:32–00:55 in Brasília). This is separate from the October 19 reports and does not establish failures in every payment method or country, or verify each transaction.
Safe next step
Check the transaction status and merchant receipt before retrying. Do not pay again solely because a notification is missing. If the result is unclear, contact the merchant or official support to avoid duplicate charges; do not reset payment equipment solely because of this resolved notice.
Needs recheck
Box API, file transfers and Sign: recovery reported
- Who / what is affected
- Content API · web uploads/downloads · Box Sign · no specific countries named
Box marked this incident Resolved at 03:47 UTC on October 20 after monitoring and reported that service functionality had been restored. The incident concerned Content API, web uploads/downloads and Box Sign. This provider recovery report does not confirm that every interrupted file transfer, API write or signature request completed.
Safe next step
Preserve local files and check whether an upload, API operation or signature request already completed before retrying. Avoid duplicate files or repeated signature invitations. Keep the error and time for official support; do not delete stored content or disable security controls as a workaround.
Needs recheck
Gemini Gateway controls and DLP incident reported resolved
- Who / what is affected
- Cloudflare Gateway application controls · DLP prompt logging and enforcement · certain Google Gemini traffic
Cloudflare explicitly marked this incident resolved at 22:46 UTC on October 20. It concerned Gateway application controls and DLP inspection for certain Google Gemini traffic, including prompt logging and enforcement. The earlier October 19 notice described a Gateway fix rolling out while DLP recovery continued; the new resolution notice supersedes that status. This is not a report of a general Gemini outage or confirmed data loss, and does not verify every organization's individual policy configuration.
Safe next step
If your organization relies on these controls, avoid submitting sensitive data through potentially affected Gemini workflows until your administrator verifies the required protections. Keep the relevant policy details and diagnostic timestamps. Do not bypass Gateway, disable DLP or relax security rules as a workaround.
Needs recheck
Support-ticket creation incident reported resolved
- Who / what is affected
- Supabase · support form and ticket creation for some customers
Supabase identified errors creating support tickets for some customers at 17:13 UTC on October 19. It reported a deployed fix and normal ticket creation at 17:36 UTC, then explicitly marked the incident resolved at 18:13 UTC after ticket creation remained normal. This concerns the support form, not a confirmed outage of existing databases or Realtime, and does not verify each individual ticket.
Safe next step
Check for a ticket confirmation before resubmitting so you do not open duplicates. Keep the error message and time, without including credentials. Do not reset a database, rotate keys or change project settings solely because of this support-form notice.
Needs recheck
October 19 Actions incident: retrospective adds impact and recovery details
- Who / what is affected
- Actions · Pages · Actions-generated webhooks · live updates and code search for part of the incident
GitHub expanded its October 19 incident report on October 22, describing impact from 15:24 to 16:52 UTC on October 19; the initial investigation notice was at 15:49. Actions status, logs and checks, Pages deployments and Actions-generated webhooks were delayed, with live updates and code search also affected for part of the period. GitHub reports no data loss. A connection retry surge overloaded a shared network link; mitigation reduced connections and moved a service away from affected cache servers. The initial connection timeouts remain under investigation. This retrospective is not a new outage or proof that each job succeeded; the separate October 22 Actions-results incident remains under investigation.
Safe next step
Check workflow run status, deployment history and webhook delivery records before retrying. Confirm whether a production deployment or downstream action already ran to avoid duplicate changes. Do not force-push code, reset repositories or rotate credentials solely because of this resolved incident.
Needs recheck
File-download error incident reported resolved
- Who / what is affected
- OpenAI · file-related ChatGPT and API components
OpenAI reported elevated errors at 13:31 UTC on October 19 and marked its File Download errors incident resolved at 14:14 UTC. The official page lists API Responses, Images and Files, plus ChatGPT file uploads and image generation. This does not establish that every request or model was unavailable, or that an interrupted file operation completed.
Safe next step
Keep any existing files and results. Check whether the operation completed before retrying, especially for API jobs that may create duplicate work or charges. Do not delete conversation data or disable browser protection solely because of this resolved notice.
Needs recheck
Two Brazil in-store QR incidents reported resolved
- Who / what is affected
- Brazil (MLB) · in-store QR payments · two separate reports
Mercado Pago published two separate reports of increased in-store QR payment declines in Brazil on October 19: 12:07–12:20 UTC and 12:57–13:00 UTC (09:07–09:20 and 09:57–10:00 in Brasília). Both were explicitly marked resolved. These are not evidence of a continuous outage between the reports, or of failures in every payment method or country.
Safe next step
Check each transaction and the merchant receipt before retrying a payment. If the result is unclear, contact the merchant or official support to avoid duplicate charges. Do not reset payment equipment solely because of these resolved notices.
Needs recheck
DNS record-update delay reported resolved
- Who / what is affected
- Cloudflare DNS record updates · propagation delays
Cloudflare investigated delays in propagating updates to DNS records from 11:28 UTC on October 19 and marked the incident resolved at 11:50 UTC. The notice concerned record updates, not a confirmed outage of all DNS lookups. Provider recovery does not prove that every resolver or client already has the latest record.
Safe next step
Compare saved record values with authoritative DNS and your deployment records. Respect the record's cache lifetime if old answers persist; do not repeatedly rewrite records or change nameservers solely because of this resolved notice.
Needs recheck
Argentina in-store QR payment incident reported resolved
- Who / what is affected
- Argentina (MLA) · in-store QR payments
Mercado Pago reported increased declines for in-store QR payments in Argentina at 10:02 UTC and marked the incident resolved at 11:50 UTC on October 19 (07:02–08:50 in Argentina). The report did not establish that every payment method or other countries were affected. A recovery notice does not verify each individual transaction.
Safe next step
Check the transaction status and merchant receipt before retrying. If the result is unclear, contact the merchant or official support to avoid duplicate charges; do not reset payment equipment solely because of this notice.
Needs recheck
Notion Agent incomplete-request incident reported resolved
- Who / what is affected
- Notion Agent · requests stopping before completion
Notion marked the issue causing some Agent requests to stop before completion resolved at 15:51 UTC on October 18 and reported normal Agent operation. This does not confirm that every interrupted task completed. The earlier real-time/API incident and planned October 20 migration remain separate.
Safe next step
Keep your prompt and any partial results. Check what the agent already changed before retrying a task to avoid duplicate work; do not clear local data or reset the workspace.
Needs recheck
Virginia and Ohio latency incident reported resolved
- Who / what is affected
- United States · Ashburn (IAD) and Columbus (CMH) · October 18, 08:00–08:50 UTC
Cloudflare retrospectively reported increased latency and timeouts on external-provider peering links in IAD and CMH. It says services fully normalized after routing advertisements were restored. The reported impact window ended at 08:50 UTC; the notice was published at 09:37 UTC, not the start of a new outage.
Safe next step
Compare your error times and traffic location with the reported window. Keep request identifiers for support if problems continue; do not change DNS or redeploy solely because of this resolved report.
Needs recheck
Uruguay Point payment-rejection incident resolved
- Who / what is affected
- Uruguay (MLU) · Point payments
Mercado Pago marked the Point payment-rejection incident in Uruguay resolved at 02:57 UTC on October 18 (23:57 in Uruguay on October 17). The initial report was posted at 01:38 UTC. This incident did not establish that all payment methods or countries were affected; the recovery notice does not verify each individual transaction.
Safe next step
Check the transaction status and merchant receipt before retrying. If the result is unclear, contact the merchant or official support to avoid duplicate charges; do not reset the terminal solely because of this notice.
Needs recheck
Supabase project creation and unpausing restored in ap-northeast-1
- Who / what is affected
- ap-northeast-1 · creating projects and resuming paused projects
Supabase reported errors when creating or unpausing projects in this region during an upstream provider outage. At 07:49 UTC on October 17, it confirmed full restoration and re-enabled both operations. This notice does not establish that all running databases or all regions were unavailable.
Safe next step
Check the project's current state before retrying creation or unpausing. Keep error details for support if it still fails; do not delete or recreate a project, or move regions, solely because of this resolved incident.
Needs recheck
Netlify build errors affecting the CSP extension resolved
- Who / what is affected
- Build Pipeline · builds using the CSP extension
Netlify identified a third-party service issue affecting builds that use the CSP extension. It released a fix, monitored the results and marked the incident resolved at 05:55 UTC on October 17. This report does not establish an outage of all hosted sites.
Safe next step
Retry an affected build normally and check the build log. If errors continue, save the build ID and error for Netlify support; do not remove security headers or disable CSP as a workaround.
Needs recheck
October 17 IPN and Webhook notification delays resolved
- Who / what is affected
- Payment notifications · all sites in the provider notices
Mercado Pago reported new IPN and Webhook payment notification delays on October 17. It marked Webhook notifications resolved at 03:50 UTC and IPN at 04:00 UTC. These are separate from the October 15–16 incidents; a delayed notification does not by itself mean a payment failed.
Safe next step
Check the transaction in the official account before retrying a payment or fulfilling an order. Reconcile delayed notifications with transaction records to avoid duplicate charges or fulfillment.
Needs recheck
Cloudflare FedRAMP Dashboard and Access incident resolved
- Who / what is affected
- FedRAMP Dashboard · applications protected by FedRAMP Access
Cloudflare marked this incident resolved at 19:09 UTC on October 16. Its follow-up included login failures in the FedRAMP Dashboard and applications protected by FedRAMP Access. Cached-file delivery through the CDN and other Edge security features were not affected by this incident.
Safe next step
Retry the affected dashboard or protected application normally. If access still fails, keep the error and time for your IT team or official support; do not change DNS or disable access controls.
Needs recheck
Later October 16 IPN notification delay resolved
- Who / what is affected
- IPN payment notifications · all sites in the provider notice
Mercado Pago reported another IPN notification delay at 13:45 UTC and marked it resolved at 14:01 UTC on October 16. This is a separate incident from the earlier IPN/Webhook reports; a delayed notification does not by itself prove that a payment failed.
Safe next step
Check the transaction in the official account before retrying a payment or fulfilling an order. Reconcile delayed notifications with transaction records to avoid duplicate charges or fulfillment.
Needs recheck
Brazil in-store QR payment incident resolved
- Who / what is affected
- Brazil (MLB) · in-store QR payments
Mercado Pago reported increased declines for in-store QR payments in Brazil and marked the incident resolved at 13:16 UTC on October 16. This was separate from the Uruguay QR and Chile Point incidents; it was not a report of all payment methods failing.
Safe next step
Check the transaction status and merchant receipt before trying again. Do not repeat a payment solely because its notification is missing; contact official support if the result remains unclear.
Needs recheck
Chile Point payment-approval incident resolved
- Who / what is affected
- Chile (MLC) · Point payments
Mercado Pago reported a payment-approval anomaly affecting Point in Chile and marked it resolved at 11:49 UTC on October 16. The report does not establish a worldwide payment outage.
Safe next step
Check the transaction status and receipt before retrying a payment. If the result is unclear, contact the merchant or official support to avoid duplicate charges.
Needs recheck
Cloudflare Dashboard signup issue resolved
- Who / what is affected
- Dashboard · account signup for some users
Cloudflare reported that some users could not sign up to its Dashboard. The incident was marked resolved at 11:24 UTC on October 16; this report was about signup, not a general website-delivery outage.
Safe next step
Retry signup through the official Dashboard. If an account may already exist, use its sign-in or recovery flow before creating another account; no DNS change is needed for this signup issue.
Needs recheck
Cloudflare Access post-login issue resolved
- Who / what is affected
- Cloudflare Access · after login
Cloudflare investigated problems with Access after login and marked the incident resolved at 09:59 UTC on October 16. The notice did not specify affected countries or establish a general Internet outage.
Safe next step
Retry the protected application normally. If access still fails, give your IT administrator the error and time; do not disable Access policies or bypass your organization's controls.
Needs recheck
Uruguay in-store QR payment incident resolved
- Who / what is affected
- Uruguay (MLU) · in-store QR payments
Mercado Pago reported increased declines for in-store QR payments in Uruguay and marked the incident resolved at 04:51 UTC on October 16. This was a regional payment issue, not a worldwide outage.
Safe next step
Check the payment status and merchant receipt before retrying. Do not pay again solely because a notification is missing; avoid duplicate charges.
Needs recheck
October 16 IPN and Webhook notification delays resolved
- Who / what is affected
- Payment notifications · provider labels both reports all sites
Mercado Pago marked separate October 16 notification delays resolved: Webhooks at 03:42 UTC and IPN at 03:53 UTC. These are new reports, separate from the October 15 IPN incident; notification delays alone do not establish failed payments.
Safe next step
Reconcile notifications against payment records before retrying or marking orders unpaid. Use duplicate-safe processing and do not automatically repeat a charge.
Needs recheck
Workers Cron Triggers incident reported resolved
- Who / what is affected
- Workers Cron Triggers · a subset of customers
Cloudflare reports that Workers Cron Triggers have stabilized and marks this incident resolved. The earlier missed scheduled triggers affected a subset of customers; check your own jobs for missed or duplicated work.
Safe next step
Check scheduled-run history and logs before manually rerunning a job. Confirm whether previous work completed and use duplicate-safe retries; do not replay payments or other irreversible tasks blindly.
Needs recheck
Payment notification delays
- Who / what is affected
- IPN payment notifications · provider labels all sites
Mercado Pago has marked this IPN notification incident resolved. The earlier delay did not establish that payments themselves had failed; individual transactions still need to be checked separately.
Safe next step
Verify the individual payment in the official account or authorized integration before retrying a charge or fulfilling an order. A missing notification is not proof of non-payment.
Needs recheck
Latency and errors at two eastern North American facilities
- Who / what is affected
- Traffic handled by Ashburn (IAD) and Newark (EWR)
Cloudflare has marked the network performance incident at Ashburn and Newark resolved. The report concerned traffic handled by these facilities, not a worldwide outage.
Safe next step
Compare error times and the affected traffic path with the official notice. Preserve request identifiers for support; avoid unrelated DNS changes or repeated deployments.
Needs recheck
Durable Objects errors in DEN
- Who / what is affected
- DEN · Durable Objects and dependent services
Cloudflare has explicitly marked the DEN Durable Objects incident resolved. The reported errors affected Durable Objects and dependent services in DEN, not every dependent service worldwide.
Safe next step
Check whether the failing operation uses the affected service and location. Keep error identifiers and verify prior writes before retrying operations that could create duplicates.
Needs recheck
pages.dev access issue in Norway
- Who / what is affected
- Norway · pages.dev subdomains
Cloudflare has marked this pages.dev access incident in Norway resolved. Custom domains were not affected by this report.
Safe next step
Compare the visitor’s region and hostname with this notice. Avoid unrelated DNS changes; check the official update before retrying deployment.
Needs recheck
Compliance API cost-data delay reported resolved
- Who / what is affected
- Organization cost reporting · not ChatGPT availability
OpenAI has marked the Compliance API cost-data delay resolved. This incident concerned organization cost reporting, not API request processing or ChatGPT availability.
Safe next step
Recheck the affected reporting periods and reconcile totals in the official account. A provider recovery notice does not validate every individual bill or report.
Needs recheck
Infrastructure migration completed; incident reported resolved
- Who / what is affected
- Real-time updates · API/MCP · Notion AI screenshots · agents · workspace members
At 18:56 UTC on October 21, Notion explicitly confirmed that its infrastructure migration had completed without customer impact, service remained stable and this incident was resolved. This supersedes the earlier migration-in-progress notice covering real-time updates, API/MCP, Notion AI screenshots, agents and workspace members. It does not erase the earlier disruptions or prove that each queued change or agent run completed; the report time is not an exact migration-completion time.
Safe next step
Preserve unsynced work. Confirm whether a write reached the workspace before repeating it; avoid clearing local data while edits are pending.
Needs recheck
Realtime latency incident reported resolved
- Who / what is affected
- Realtime presence and channel joins
Supabase marked this incident resolved. It was not a report of a complete database outage.
Safe next step
If your app still fails, check its subscriptions and logs separately. A resolved provider incident does not prove every client has recovered.
Check the match before you change anything
1 · Match the symptom
Compare the exact error, service feature, country, app version and timing. A nearby incident may be unrelated to your problem.
2 · Protect your work
Keep a copy of unsynced work and note the error and time. Avoid repeated payment attempts, destructive resets or security changes while the cause is unclear.
3 · Choose the next check
Use the live monitor for provider reports. If there is no matching report, check a second device or network only where safe and permitted; absence of a report is not proof of normal service.
Keep checking
How this desk works
We select official provider status notices, release-health records and release notes, then distinguish reported impact from our suggested next step. We do not infer global downtime from one region, a resolved incident from a missing feed entry, or device recovery from an available update. Public reports cannot confirm the health of your particular account or tenant.
The live monitor polls official sources separately, typically every 5 minutes where supported. These pages add context on a 3-hour review target. Collection failures and overdue reviews are not reported as healthy service.
Coverage starts with selected cloud, collaboration, Windows and browser issues relevant to WiseChecker readers. This is not a complete list of every service, region or security vulnerability. For urgent decisions, use the provider’s current notice and your organization’s guidance.
Official links may be in English; the summaries and guidance here are localized.