supabase / storage / 42819457742

88%

push

github

fix: suppress repeat migration submissions with cooldown (#1685)

## What kind of change does this PR introduce?

Bug fix

## What is the current behavior?

Once a tenant batch was submitted, another request could submit it
again. pg-boss deduplicated outstanding jobs but each submission still
reached the queue database.
Previously added backoff and jitter applied when the batch send failed.

## What is the new behavior?

After a successful batch submission, the process suppresses repeat
submissions for that tenant for 60–120 seconds with jitter. The cache
holds up to 4096 tenants by default and both settings are configurable
via env.
Failed sends retain their existing retry behavior and admin migration
fallbacks bypass the cooldown.

## Additional context

The cooldown is per process so it reduces load but is not a fleet wide
limit and a new request after expiry can submit again.

Signed-off-by: Ferhat Elmas <elmas.ferhat@gmail.com>

8376 of 10080 branches covered (83.1%)

Branch coverage included in aggregate %.

16 of 17 new or added lines in 3 files covered. (94.12%)

14354 of 15902 relevant lines covered (90.27%)

1301.02 hits per line