Thank you, Matt. There is indeed some good, detailed technical information hidden in an attachment there...
Unfortunately my timestamp observations are still partly unexplained:
In a loop with only an FP-read on LED-A and a wait-on-next-millisecond-multiple running on the cFP-2020 I would expect the difference between consecutive timestamps (deltaT) to be very close to the wait time. My results:
wait 1000 ms: deltaT= 0 or 1000 or 2000
wait 5000 ms, deltaT= 4000 or 5000 or 6000
wait 333 ms, deltaT= 0 or 333 or 1333
(there's also a random variation of a few milliseconds on top of this, which I'm not concerned about)
How can this be?