Secondly the fact that various data sheets have no mention about what happens in writing these bits means that you have to check your own derivative to see how it actually works.

oh, they do, the only thing that could be questionable (could be, I do not think so) is whether you can write to the bits from code.

The fact that e.g. TF1 causes the interrupt is 'biblical' and I have seen no exception to the methodology of synchronozation: the event set a SFR bit, the SFR bit causes the interrupt synchronously.

Erik