From your docs:
“If a container is successfully restarted (the container is started and runs for at least 10 seconds), the delay is reset to its default value of 100 ms.”
10 seconds may like a sound reasonable (or even too long) timeout for many applications, but it is not enough for non-small databases. Our production MariaDB databases take 15 or even more seconds to initialize the buffer pool. So, MariaDB could fail after more than 10 seconds because a file is corrupted, or could be killed by the OOM after 10 seconds (but before finishing the initialization).
My request is: please, make the 10 seconds timeout configurable on a per-container basis.
(this FR was originally filed on GitHub as #15623, now I closed it because it isn’t a bug report)