fix(traefik): log all 4xx/5xx access events, not just slow retries

Traefik combines accessLog.filters fields with AND, so configuring
both `retryAttempts: true` and `minDuration: '1s'` alongside
`statusCodes: 400-599` meant a request had to be a 4xx/5xx AND
have been retried AND have taken longer than 1s before it would
appear in the access log. Most production errors were silently
dropped — including the fast 500s that matter most.

Drop the two extra filters so any 4xx/5xx is logged. Volume stays
manageable because errors are a small fraction of total traffic;
verbose normal-traffic logging remains gated behind the upstream
`log.level: WARN` setting.

Made-with: Cursor

Author: HMarzban

scripts/traefik/traefik.yml

@@ -24,14 +24,14 @@ log:
   level: WARN # Reduced from INFO to reduce log volume
   format: json
 
-# Access logs - only log errors (4xx/5xx) to reduce volume
+# Access logs — 4xx/5xx only (keeps volume low but still captures fast 500s).
+# Traefik combines filter fields with AND; retryAttempts + minDuration meant
+# most errors never appeared in logs. See docker-compose.prod.yml "Debugging".
 accessLog:
   format: json
   filters:
     statusCodes:
       - '400-599'
-    retryAttempts: true # Only log retries
-    minDuration: '1s' # Only log slow requests (>1s)
 
 # -----------------------------------------------------------------------------
 # Entrypoints