Nutshell consumes Postgres connection pool on low activity

[minibits-cash]

I see from time-to-time nutshell to consume all connections to Postgresql database (FATAL: sorry, too many clients already with max_connections = 100 setting) despite relatively low overall load. Connections consumed are idle, max 1 is active. Nutshell itself runs with around 12 system processes spawned.

[a1denvalu3]

Investigating.

[a1denvalu3]

@minibits-cash you don't see these errors, right?

2026-02-24 10:46:27.41 | ERROR | Failed to set proofs pending: failed to acquire database lock on proofs_pending after 1s and 4 trials (1771926386410)
2026-02-24 10:46:27.41 | ERROR | Exception: failed to acquire database lock on proofs_pending after 1s and 4 trials (1771926386410)

[minibits-cash]

I do see these around, it might trigger the FATAL .... situation, althought there is some delay. There are then like 100 lines of too many... linesprinted within a second or so.

2026-02-23 02:00:56.712 UTC FATAL: sorry, too many clients already
2026-02-23 01:59:22.684 UTC ERROR: could not obtain lock on relation "proofs_pending"
2026-02-23 01:59:22.684 UTC STATEMENT: LOCK TABLE proofs_pending IN EXCLUSIVE MODE NOWAIT;

as well as

2026-02-23 02:20:32.323 UTC FATAL: sorry, too many clients already
2026-02-23 02:04:48.685 UTC ERROR: could not obtain lock on relation "mint_quotes"
2026-02-23 02:04:48.685 UTC STATEMENT: LOCK TABLE mint_quotes IN EXCLUSIVE MODE NOWAIT;