Is it really necessary to debounce matrix keypad for normal applications?

Yes, of course! The mistake in your theory is, that debouncing is a chaotic process, which cannot be predicted in any way. This also includes the bouncing time. What you observed with your microswitch is a 'friendly' bouncing, which only lasts for a few miliseconds. To tell the truth, if you take a brand new switch you will hardly observe bouncing which lasts longer than 5msec.

BUT: This is only true for a NEW switch! The older your switch gets the less proper is the surface of contacts and the longer bouncing lasts! Think of dust penetrating the switch enclosure and contaminating the contacts. Not to forget chemicals like SO2 etc, which cause the contacts to become oxidized and sulfurized. Think also of the mechancial stress of thousands of times hitting the switch, and the fine arcs between contacts, when the switch is released.

So, your theory might be true for a new and perfect switch, but for a normal one, which is a bit aged your theory fails. From experience I know, that bouncing can take place for about 50msec. Means, if debouncing time constant is 50msec, then probability that a bouncing could not be suppressed is acceptable low.

Kai