Third-party dependency fails
Payment provider, maps, identity or model provider degrades and the product does not recover.
Severity: outage. Attached to 13 of 46 published product types, backed by 18 evidence records.
Product types that carry it
| Product type | Grade | Note for this type | Evidence |
|---|---|---|---|
| mobile banking app | primary | A migration to an outsourced core banking platform goes live and customers cannot log in, see accounts or move money in the bank's app. | |
| neobank | primary | The banking-as-a-service provider between the app and its partner banks fails, and app customers cannot withdraw balances the app still displays. | |
| accounting and invoicing SaaS | practitioner single incident, no second source | The open banking provider goes down, and bank feed imports stop until it recovers. | |
| clinical trial management software | primary | The hosted randomisation service is down during a site visit and the investigator cannot randomise or emergency-unblind a participant, with no documented fallback. | |
| wearable companion app | practitioner single incident, no second source | The vendor cloud goes down and the companion app cannot sign in or sync, although the watch on the user's wrist still holds every workout. | |
| ecommerce store | practitioner single incident, no second source | The CDN in front of the storefront returns errors and the catalogue, cart and checkout all go dark together, with no origin fallback. | |
| identity and access management service | practitioner single incident, no second source | A quota or database layer under the token-validation service fails, and SSO sign-in errors at once in every relying app of every tenant. | |
| SDK or client library | primary | A change in a server response crashes host apps at launch inside the SDK, or an unpublished transitive package makes host installs and builds fail. | |
| data pipeline | practitioner single incident, no second source | The managed stream the pipeline reads from stalls after a provider capacity change, and every job consuming it stops loading until the provider recovers. | |
| LLM-based assistant | primary | The model provider's API times out or silently returns degraded answers, and the assistant shows a spinner or a confident wrong reply with no fallback message. | |
| smart home or IoT app | primary | The vendor cloud goes down and every bulb, plug and thermostat stops answering the app, because commands only route through the cloud. | |
| EV charging app | primary | The roaming partner's network or the charger's backend link drops, and the app cannot start, stop or bill a charger it still shows as available. | |
| messaging app | practitioner single incident, no second source | A cloud network gateway in front of the chat backend fails to scale with reconnect traffic. |
Tests that address it
- integration tests: mandatory for this risk on 6 product types.
- resilience and failure testing: mandatory for this risk on 13 product types.
- release smoke tests: mandatory for this risk on 2 product types.
- device and browser compatibility: mandatory for this risk on 1 product types.
- unit tests: mandatory for this risk on 1 product types.
- contract tests: mandatory for this risk on 1 product types.
- regression suite: mandatory for this risk on 1 product types.
Evidence
ev-0027regulatorprimary
PS21/3: Building operational resilience, Financial Conduct Authority
The FCA requires UK firms to identify important business services, set impact tolerances and test that they can stay within them during severe but plausible disruption.
published undated · verified 2026-10-07
ev-0028regulatorprimary
Regulation (EU) 2022/2554 (Digital Operational Resilience Act), EUR-Lex, European Union
DORA requires EU financial entities to test digital operational resilience, manage ICT third-party risk and report major ICT incidents.
published undated · verified 2026-10-07
ev-0040post-mortemprimary
UK Open Banking customers unable to create or manually renew bank feeds, Xero
Supports dependency-failure and data-corruption for accounting SaaS: the vendor's own incident report says an open banking provider outage stopped bank feed imports, and transactions imported after recovery left duplicates that customers had to remove by hand before reconciling.
published 2022-12-19 · verified 2026-10-07
ev-0056standardprimary
ICH E6(R3) Guideline for Good Clinical Practice, Step 5, International Council for Harmonisation, published by the European Medicines Agency
Contingency procedures should be in place to prevent loss or lack of accessibility to data essential to participant safety, trial decisions or trial outcomes.
The GCP standard requires trial systems to survive system failure without losing or locking away safety and outcome data, which makes a hosted EDC or IRT outage a trial-level risk.
published 2026-07-14 · verified 2026-10-07
ev-0075post-mortemprimary
Summary of June 8 outage, Fastly
Official CDN post-mortem: a valid customer configuration change triggered a software bug and the network returned errors for customer sites. Supports dependency-failure for content management systems that deliver published content through a CDN, and for any storefront whose catalogue, cart and checkout are delivered through a CDN.
published 2021-06-08 · verified 2026-10-07
ev-0112court-or-fineprimary
TSB fined [amount] for operational resilience failings, Bank of England (Prudential Regulation Authority)
This resulted in significant disruption to the continuity of TSB's banking services, including branch, telephone, online and mobile banking.
Joint PRA and FCA penalty after a platform migration left customers unable to use mobile and online banking and the bank failed to manage risk from its critical third-party IT supplier. Supports dependency-failure and regulatory-penalty for a bank's mobile app; transfers to neobanks that run on an outsourced core banking platform.
published 2022-12-20 · verified 2026-10-07
ev-0120court-or-fineprimary
Synapse Financial Technologies, Inc. (CFPB enforcement action), Consumer Financial Protection Bureau
CFPB complaint that a banking-as-a-service middleware provider did not keep records of where consumer funds sat or keep them matched with partner bank records, so consumers of fintech apps lost access to funds. Supports data-corruption and dependency-failure for neobanks that hold balances through partner banks.
published undated · verified 2026-10-07
ev-0190post-mortemprimary
Garmin issues statement on recent outage, Garmin
many of our online services were interrupted including website functions, customer support, customer facing applications, and company communications.
A wearable maker's own statement on a cyber attack that took its cloud and companion services offline while watches kept working, which is the device-works-but-app-cannot-sync failure this product type must handle.
published 2020-07-27 · verified 2026-10-07
ev-0204post-mortemprimary
Registry operator incident report: an author unpublishing a small package made installs and builds fail for every project that resolved it. Supports dependency-failure for command-line tools installed from a public registry. The unpublished package was a library that other projects resolved at install, so it supports dependency-failure for an SDK whose transitive dependencies fail host installs and builds.
published undated · verified 2026-10-07
ev-0214post-mortemprimary
Summary of the Amazon Kinesis Event in the Northern Virginia (US-EAST-1) Region, Amazon Web Services
Amazon Cognito uses Kinesis Data Streams to collect and analyze API access patterns.
Official post-mortem: a capacity addition pushed the streaming service front-end past an operating system thread limit, and services that consume the stream were impaired. Supports dependency-failure for data pipelines built on a managed stream.
published 2020-11-25 · verified 2026-10-07
ev-0235regulatorprimary
23 CFR Part 680: National Electric Vehicle Infrastructure Standards and Requirements, U.S. Federal Highway Administration (eCFR)
The price for charging must be displayed prior to initiating a charging transaction and be based on the price for electricity to charge in $/kWh.
The US rule for federally funded chargers requires price shown before a session starts, payment methods accessible to people with disabilities, OCPP between charger and network, network-to-network roaming and real-time port status for third-party apps.
published undated · verified 2026-10-07
ev-0237regulatorprimary
The Public Charge Point Regulations 2023, regulation 6: Payment roaming, UK Government (legislation.gov.uk)
Operators must let drivers pay through a third-party roaming provider, so a charging app depends on another company's network to start and bill sessions.
published undated · verified 2026-10-07
ev-0252post-mortemprimary
Google Cloud Infrastructure Components incident (User ID Service authentication outage), Google Cloud
The Google User ID Service maintains a unique identifier for every account and handles authentication credentials for OAuth tokens and cookies.
Official incident report: an automated quota system cut the quota of the account database behind the authentication service after a migration reported zero usage, and authentication lookups failed across dependent services. Supports dependency-failure for identity services built on internal platform layers.
published undated · verified 2026-10-07
ev-0257standardprimary
ETSI EN 303 645 V3.1.3: Cyber Security for Consumer Internet of Things: Baseline Requirements, ETSI
Consumer IoT devices should remain operating and locally functional in the case of a loss of network access and should recover cleanly in the case of restoration of a loss of power.
The consumer IoT baseline standard asks for resilience to network and power outages, unique per-device passwords and a secure update mechanism, which the phone app controlling the device has to respect.
published undated · verified 2026-10-07
ev-0277post-mortemprimary
API, ChatGPT & Sora Facing Issues (incident of December 11, 2024), OpenAI
The model provider reports an outage of its API caused by a telemetry deployment that overwhelmed its cluster control plane, so assistants built on that API lost service.
published 2024-12-11 · verified 2026-10-07
ev-0278post-mortemprimary
A postmortem of three recent issues, Anthropic
This is a technical report on three bugs that intermittently degraded responses from Claude.
The model provider reports infrastructure bugs that degraded answer quality for some requests without an outage, so an assistant built on a hosted model can give worse answers with no error to detect.
published 2025-09-17 · verified 2026-10-07
ev-0286post-mortemprimary
Slack's Outage on January 4th 2021, Slack Engineering
On January 4th, one of our Transit Gateways became overloaded. The TGWs are managed by AWS and are intended to scale transparently to us.
Supports outage-under-load and dependency-failure for a messaging app: clients returning after the holidays reconnected with cold caches, and a cloud-managed network gateway did not scale with the surge.
published 2021-02-01 · verified 2026-10-07
ev-0299practitionerpractitioner
Suddenly FBSDKRestrictiveDataFilterManager.m - Line 71 started to crash, GitHub (facebook-ios-sdk issue tracker)
It suddenly started to crash from 14:44 PST 5/6/2020.
Named developer report: host apps crashed at launch inside the SDK after a server response changed, with no app release on either side. Supports dependency-failure for mobile SDKs that read remote configuration.
published 2020-05-06 · verified 2026-10-07