A hosted Bijection deployment stops itself when it detects a condition it cannot safely continue through. Stopping is the right call. Staying stopped is not.
Restart in place
The deployment now restarts in place after such a stop, on the same machine and with the same configuration. There is no new instance to provision and no address to change.
Wait for the writer
On startup, the deployment waits until its database accepts writes before it serves anything. A process that is running but cannot commit is not ready, and it no longer pretends to be.
What you will see
Recovery that used to need an operator now completes on its own. The deployment's activity log records the stop and the restart.