Fixed
- The watchdog restarted the server after every wake from standby. Standby suspends the watchdog along with everything else, so on resume it saw a heartbeat that had stopped when the whole TV did and treated it as a wedged event loop. It now times its own loop and, where far more time has passed than it slept for, gives the server one round to write a beat before judging it.
A genuine wedge is still caught — one check later at worst. Verified on a B8 by freezing the server and the watchdog and resuming only the watchdog, which reproduces the resume ordering standby produces:
16:40:37: resumed after 147s suspended - skipping one check
16:41:07: heartbeat 167s old - event loop wedged, restarting
Fixes the known issue noted in v0.6.0.