Anomalies connues
Suivi des problèmes identifiés et de leur résolution
Analytics showed a brief data gap during migration
During the database migration, real-time analytics went dark for a few hours. No data was lost. The gap only affected live visibility, not historical records.
We backfilled all historical data (260M+ records) and verified per-day consistency. Real-time ingestion came back within hours.
Safari users hit a login redirect loop on the dashboard
If you used Safari, logging into the dashboard triggered a redirect loop. A WebKit cookie handling quirk caused the session to fail across subdomains.
We fixed cookie attributes for cross-subdomain compatibility. Login works correctly on all major browsers now.
Stale proxies stayed in the pool during peak hours
During peak usage, the proxy validation pipeline couldn't keep up. Some requests hit recently-failed proxies before the pool evicted them.
We added parallel health checks and faster dead-proxy eviction. The pool now stays fresh even under 10x normal load.
Browser requests timed out under heavy JS loads
Browser-mode requests were timing out on pages with heavy JavaScript bundles. Headless Chrome didn't have enough shared memory to handle concurrent sessions.
We increased shared memory for browser sessions and added per-request resource limits. Timeout rate dropped from ~8% to under 0.5%.