Persist error data across reboots
If the watchdog fires and the NMI handler is called, a range of error parameters are collected together in the platform code. However, we cannot attach a debugger at this point in our field deployed products, so we cannot get the error information.
We would like to take the error info and store it somewhere; then do a reboot; then access the error data again so it can be sent off to our host in a message (we do not feel we can still send a message reliably in the NMI handler itself).