Service outage01
INCIDENT / 2026-10-08
A reverted deployment can leave interacting configuration behind. For storage-dependent workloads, define which routing, planner and metadata states must return to baseline, and who confirms each one. A lower aggregate error rate is not enough if a small backend cohort still receives disproportionate traffic. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage02
INCIDENT / 1 September 2026
A regional network incident raises a practical buying question: do your recovery plans depend on infrastructure that can fail together? Read Google’s account and use the lessons to challenge assumptions about isolation and recovery.
Iowa, United Statesvendor asserted
Service outage03
INCIDENT / 2026-08-20
Restored connectivity does not establish that queued work has completed or lifecycle operations are healthy. Record separate recovery milestones for incoming requests, accumulated jobs and administrative changes. A regional exit plan also needs available destination capacity and working identity dependencies. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage04
INCIDENT / 2026-07-15
Two feeds or redundant components do not alone demonstrate independent recovery. Map the controllers, transfer logic, maintenance exceptions and personnel needed for the alternate path. Evaluate the service you actually buy; one facility event can produce different recovery milestones for different products. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage05
INCIDENT / 2026-07-14
A green routing session is evidence about that session, not every tunneled workload. Require observability for the actual private address space and inter-zone data exchange. Keep configuration syntax validation separate from behavioral validation under the deployed topology. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage06
INCIDENT / 2026-06-05
An alternate route can be reachable yet too constrained for the workload. Procurement evidence should distinguish route availability, peak sustainable throughput and latency from actual user locations. Do not assume a cloud region outage from a non-compute network facility event. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
AI quality07
INCIDENT / March–April 2026
Anthropic’s quality postmortem describes changes around the model rather than a failure of its API inference layer. The lesson for evaluators is to preserve prompts, settings, session state and application versions alongside the model name.
Geography not establishedvendor asserted
Service outage08
INCIDENT / 2026-02-27
Switching model names is not proof of an independent failure path. Ask whether fallback routes share filtering, identity, quota, region or billing dependencies. Error handling also needs to distinguish temporary capacity failures from policy denials without bypassing the application’s safety requirements. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage09
INCIDENT / 2026-02-20
Stopping a cleanup task does not necessarily restore the objects and service bindings it already removed. Before automating bulk deletion, bound the eligible set, define a maximum change count and retain a recoverable configuration snapshot. Separate reachability restoration from completion of the recovery workflow. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage10
INCIDENT / 2026-01-22
Removing a restrictive condition can widen a policy even when the change looks like cleanup. Review a policy’s evaluated result for representative allowed and denied cases. The same reasoning applies to data access filters and supplier-managed routing changes, without assuming identical technical mechanisms. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage11
INCIDENT / 2026-01-08
A producer-side optimization can preserve intended meaning while breaking a consumer’s assumptions. Ask which clients, versions, cache states and partially expired records were exercised. Avoid converting a successful modern-client test into a claim about the whole installed estate. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage12
INCIDENT / 2025-12-05
A gradual software rollout provides limited protection if a related operational flag changes everywhere immediately. Map every change channel used during mitigation, including internal testing controls. Decide the acceptable failure behavior explicitly; do not generalize this incident into advice to disable security checks. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage13
INCIDENT / 18 November 2025
A database permissions change fed into a network configuration failure, according to Cloudflare. Use the report to examine how generated configuration is validated, distributed and rolled back across your own critical dependencies.
Global network scopevendor asserted
Service outage14
INCIDENT / 2025-09-12
A user action can produce many dependent requests through renders and retries. Evaluate the fan-out and shared authorization dependency under failure, not only normal request latency. Increasing server resources may mask amplification without removing it. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Security & privacy15
INCIDENT / 2025-08-12–17 (reported exfiltration window)
A support ticket can carry configuration details or secrets beyond its apparent business purpose. Include chat-to-CRM integrations, OAuth access, ticket retention and internal comments in the supplier data-flow review. Rotation alone does not establish what was accessed or whether downstream misuse occurred. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage16
INCIDENT / 2025-08-21
Nominal aggregate bandwidth can hide reduced capacity on the alternate path. Ask for evidence of headroom when a link is already degraded and traffic shifts again. Keep provider-specific peering behavior distinct from your own application recovery assumptions. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage17
INCIDENT / 12 June 2025
Google’s incident account describes policy metadata reaching a previously unexercised failure path. The review question is broader than one outage: which shared control services can interrupt otherwise separate workloads?
Global service scopevendor asserted
Service outage18
INCIDENT / 2025-06-12
A secondary platform dependency can affect authentication and configuration across several products at once. Ask which emergency access paths depend on the same storage or identity service as production. Any alternate access must preserve scoped permissions and an audit trail. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Security & privacy19
INCIDENT / August 2022
The ICO’s final enforcement record describes an account without MFA in the access path to a healthcare subsidiary. The practical lesson is to verify coverage and exceptions for real accounts, rather than treating a policy statement as proof.
United Kingdomregulatory record
Service outage20
INCIDENT / 2025-03-21
A successful secret deployment is not proof that production consumers use it. Require an environment-specific check before revoking the previous credential, with a bounded overlap and rollback plan. Do not retain an old credential indefinitely merely to avoid testing rotation. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage21
INCIDENT / 2025-02-06
An account or abuse workflow can reach shared production infrastructure if object boundaries are unclear. Require a preview of the effective target, protected-service exclusions and independent review for exceptional actions. Separate storage integrity from the completeness of downstream logs. Use the preserved provider account below to distinguish reported facts from the proposed review exercise.
See reported service scopevendor asserted
Service outage22
INCIDENT / 19 July 2024
CrowdStrike’s root-cause announcement identifies a Windows sensor content update as the trigger for a widespread outage. Use the dated account to examine update controls, recovery access and the evidence behind continuity promises.
Geography not established in this sourcevendor asserted