So, the "unnecessary" interrupt took up 2% of your precious processing time... I don't think this is really worth even mentioning...

An often forgotten fact is that while, agreed, "2% of your precious processing time" is not much DURING the execution it is 100% of "your precious processing time". I can think of many cases (and know of some) where the fact that 100% of the processing is occupied at a random 2% of the time will lead to failures. Just a very simple example: you have some device that accumulates timer readings, when it has enough it stop reading time and ask the master to start reading data; A 2% of the time random access wuill royally screw up the timing reads, but with automatic address recognition there are no interrupts till ready and the timing will be precise. I pick an example that is simple to explain, the examples that REALLY get you in trouble are not that simple to explain.

There is a world of difference between occupying "2% of your precious processing time" in the main and occupying "100% of your precious processing time randomly 2% of the time"

Erik