I dunno off the top of my head, but:

* you've checked the errata?
* if you were using async you could use `with_timeout` to force a reset
* without, it might be with patching the blocking hal to "give up and reset" after some amount of time to prevent a lockup