As suggested by the file name, Vector64 is DMA1.
I've reduced the code down to the minimum project that does it (tested on a nucleo g031): https://github.com/mck1117/chibi_sv8
One thread continuously samples the adc, only one sample at a time (setting/clearing a pin while adcConvert is running). One thread toggles a GPIO every 3ms. I hooked up a logic analyzer to see some history of what's going on, and the exception occurs something like 10ms after startup. I also added a context switch hook to toggle a third pin to see what the scheduler is doing.
Any idea what's going on? Looking in the debugger, the priority of that interrupt is 10, so it shouldn't be able to preempt the kernel unless something has gone wrong (besides, on v6m there is no partial disable, you can only disable interrupts fully).
