FailureAtlas: most severe LLM gateway failures return HTTP 200 and silently corrupt your app
its_vayishu · x · 2026-09-28
metriqual published a paper, FailureAtlas, taxonomizing failure modes in multi-provider LLM serving infrastructure (gateways/reverse proxies). Key finding: the most operationally severe failures are silent — they return HTTP 200, pass every standard health check, and corrupt conversation state, requiring semantic-level observability to detect.
Highlights:
- Two-axis taxonomy: failures classified by origin layer (Network/Transport, Streaming/Protocol, State/Session, Model Behavior, Governance/Cost) and detectability (Loud vs. Silent).
- Five verified entries sourced from public bug reports and first-hand stress testing, each with root-cause analysis; three include standalone reproduction scripts.
- Two first-hand silent failures: a concurrency race causing history loss, and a streaming index collision corrupting tool-call payloads.
Implication: HTTP status codes and health checks alone are insufficient for production LLM features; semantic monitoring is required.
More from Infra
- Sarvam AI Launches Token Factory: India-Hosted Inference for Open Models — itsOmSarraf_ · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16
- Stratechery's Fictional Obituary: Imagining Jensen Huang's 2026 Exit From Nvidia — Gregory_C_Allen · 2026-11-16
- vLLM serves DeepSeek V4.1-Flash at 954K tokens/s per GPU, $38.96/GPU-hour on GB300 NVL72 — bookwormengr · 2026-11-16