Michael, you are bang on. Thank you.

In fact, pro users of audio equipment who are used to punching tracks in and out of record with extreme precision will dislike 50ms!! These guys demand high quality switches, and may even make measurements with a scope before they decide to buy a high value system - because these things are sometimes not apparent until you have used something for a few weeks. It's application dependent. What's OK for a lift button is not OK for other things ... and > 100 ms is definitely bad design.

(If you think this is physically impossible, consider that some musicians accurately produce chains of notes at 200ms intervals, using devices much less precise than a microswitch!)

The other point I want to make, is how can a switch possibly bounce if noone pressed it? So, in the absence of noise, why should we not believe the FIRST sample that says it was pressed?

About the noise issue - if the switch is in the same box as the uP, and the environment is not exceptionally noisy, why should expect any more spurious signals on the pin than we would for any other IO device?

When was the last time you designed, say, an I2C interface taking double samples of everything just to make sure? For a UART, yes ... but when all the signals are in the same physical enclosure? What would be the justification?