There has been some dialog on this thread with the fine folks in support (Thank you Mike!). It seems like a potential root cause for this is due to how the AP’s react, in this case to spanning-tree when connected to a Cisco Switch.
The interaction between the two gets stuck in a Listening state, and never actually goes to forwarding. The 2.0i flavor of firmware really brought this issue out in a ‘bad way’.
Temporarily rolling back to 1.1 seemed to resolve the issue enough that I could get an operational wireless network again.
Moving the ports these AP’s were connected to to ‘portfast edge’ allowed me to upgrade the AP’s back to 2.0j and they’ve been in operation for a week now without the symptoms I originally posted for.
Monitoring the environment for now.