Problem with the states Sablier fetches for the deployments

[geosaglio]

Describe the bug
Sablier sometimes returns the wrong state for the deployments of the group. In the theme that is sent to the user, all the deployments are marked as ready, but you get stuck on the page. This is due to the fact that some deployments that are in the group don't change their replica count from 0 to 1. Manually changing the replica count temporarily fixes the issue, but it still happens quite often.

Context

• Sablier version: 1.8.1
• Provider: kubernetes v1.30.5]
• Reverse proxy: traefik v2.11.4
• Sablier is running inside a container

Expected behavior

The deployments in a group should all either be up (at least 1 replica) or down (0 replicas) but that isn't the case.

Additional context
The timeout for the group is 2 hours, but there is a curl that is sent every 2 hours, could this be the cause of the unpredictable behavior?
I also get logs where a deployment is starting and stopping at the same time from sablier.

[acouvreur]

This might happen when the session duration is invalid, are you sure that the session_duration is long enough ? An invalid value may result in 0s session duration which would shut it down instantly

[geosaglio]

Hey!
Thanks for your quick reply, the session_duration is set to 2 hours and seems to be valid as I'm getting these logs at the same time: