AppService unhealthy instance does not auto-restart after one hour

I have a PremiumV3 Linux app service plan with 3 instances, and at some point in the last week, 2 instances started reporting as unhealthy. Since health check is enabled, the instances were removed from rotation as expected, but I also expected the health check to replace the unhealthy instances with new ones after one hour.
This did not happen and the instances remained unhealthy until I did a manual restart of the app service, which restarted all 3 containers.
Does anyone have a hint on why the health check couldn't do an "auto-restart" or it couldn't replace the instances? They remained unhealthy for up to 6 hours before they were manually restarted.
P.S: Configuration is in place to export logs to a dedicated LogAnalytics workspace, but I cannot find any events related to the auto-restart not being triggered / failing to be triggered for the appservice.