This pull request has been mentioned on openHAB Community. There might be relevant details there:

https://community.openhab.org/t/homekit-stopps-working-after-a-certain-time-7-days/161003/11

Not sure if this is the right fix. While it would be an improvement for some specific cases. It actually hides the real problem.

1. Wouldn’t it be better to solve the unstable connection itself?
2. If the connection is broken due to whatever reason, the HomeKit binding should be able to handle it. Independent of being down 1 or 50 times a day. Obviously there is a limit to this, but I would assume that when HomeKit binding reaches this limit the whole host is unusable.

agree. it is not a proper solution, more a temporary workaround for a very strange issue.
it is still unclear to me what to fix.

my current theory is that VMs get very quick interface down & up, in few milliseconds. and either homekit addon (or OH) is messing up the sequence and first adding homekit to mDNS and then removing it, or mDNS is messing up the sequence.

it is difficult to troubleshoot as i cannot reproduce this issue on my side. i can interrupt network connection as often as i wish, homekit gets restarted correctly.
my hope, with this switch we could at least confirm that the issue is related to network interface changes.

Could it be that it is so short in time that the port is still occupied. Not sure if that makes sense, but the idea came to mind when reading this:
#17998 (comment)