roject: contextos-backend (id: 9d1fd8bc-bda6-45cb-8082-3e3455694b89)

Service: Postgres (id: bddc3d20-4b75-4d06-a581-995bf74bfdf0)

Environment: production

Postgres on this service is rejecting all external connections with FATAL: password authentication failed for user "postgres". Diagnosis so far:

POSTGRES_PASSWORD in Variables tab matches the password component embedded in DATABASE_PUBLIC_URL. Railway's published credentials are internally consistent.

Connections worked normally earlier today, then started failing around [insert your approximate UTC timestamp of first failure]. No password rotation was performed on my side.

Redeployed the Postgres service (deployment id: [grab from the latest deployment row]). Auth still fails with the same error.

The TCP proxy at shinkansen.proxy.rlwy.net:47600 is reachable and responsive (Postgres returns FATAL, not a network error), so this is specifically a credential rejection by the Postgres pod, not a proxy or network issue.

The dashboard's own Database tab can read tables fine, so an admin-level connection path exists internally.

Hypothesis: the password stored in pg_authid on the persistent volume has diverged from POSTGRES_PASSWORD env var. Standard Postgres docker behavior is to only honor POSTGRES_PASSWORD during initdb on first volume mount, so subsequent env var changes don't update the stored password.

Requested action: either reset the stored Postgres credentials to match the env var, or provide a path for me to run ALTER USER postgres WITH PASSWORD '...' against this instance.