fix: replace the backend health poller with a connectivity state machine (no more false Server Unreachable during recordings)#126

Merged

Valtora merged 3 commits into

mainfrom

fix/connectivity-state-machine

Jul 24, 2026

Commits on Jul 24, 2026

fix(frontend): replace the fixed-interval backend health poller with a passive-first connectivity state machine that derives reachability from real request outcomes (via the axios interceptor), bro…
Valtora
committed
fix(api): exempt the /health and /api/health liveness endpoints from EnforceCanonicalHttpsMiddleware so container health checks and plain-HTTP uptime monitors get a clean 200 instead of a canonical…
Valtora
committed
fix(deploy): correct the nginx HTTP to HTTPS redirect that pinned the HTTPS target to the plain-HTTP port ($server_port resolves to 80 on the listen 80 block), producing a malformed https://$host:8…
Valtora
committed