When upgrading, all running tor instances are stopped (not restarted, as
expected)
syslog shows:
Interrupt: we have stopped accepting new connections, and will shut down
in 30 seconds. Interrupt again to exit now.
Clean shutdown finished. Exiting.
(problem is reproducible)
I just had the same experience upgrading my relays, but I think this is to be expected? New connections are blocked and there's 30 seconds to give existing connections a chance to gracefully complete. The daemon is then stopped while the packages upgrade, then it's restarted. I think it's been handled like that for a while, although my memory is a little fuzzy since I hadn't upgraded in the last 6 months.