Just a note to comment on the resolution of the problem. The extraordinary
scheduling latency I had problems with was caused an application error,
rather than by the processor SMM. It was due to an unused timer running with
no consumer for its pulses. Apparently that causes the kernel to traverse
through some long lists. I’ve seen mistakes like that in the past that
generate a “work-panic” type error from the kernel. It seems QNX is happy to
continue no matter how long it takes to work through a list. Oh well, live
Regards and thanks to all for your various suggestions.