hello, greater by a factor of 100 than what?

I compared the source impedance of 'key + pull-up' combination with the source impedance of standard 74HCMOS push-pull output: If you connect a 4k7 pull-up resistor at certain input of micro, then probability that false trigger occurs is 100 times greater then the probability that a false trigger occurs, if the same input of micro is fed by a 74HCMOS push-pull output!

I suppose it also deponds on the actual value of stray capacitace, dosn't it?

Yes, of course. But don't underestimate this stray capacitance! 15cm long copper trace presents an intrinsic space-charge capacitance of about 27pF! In combination with 4.7kOhm pull-up resistor a first order high pass filter is created showing a corner frequency of

fc = 1 / 2 / pi / 27pF / 4700Ohm = 1.3MHz

Means, interference of frequencies higher than 1.3MHz cannot be attenuated by the impedance of stray capacitance, but is fully injected into the input of micro!
Now consider an interference showing relevant spectral density arround or above 1MHz, think for instance of an ESD event near the keyboard, a heavy triac pulse, a fast switched-off inductive load, whatever you want that has enough energy about 1MHz, then there's a relevant probability of occuring false triggers.

Would this 15cm long copper trace be connected to the push-pull output of a 74HCMOS gate, on the other hand, then interference would have to be 100 times stronger, to cause a comparable probability of occuring false triggers!

By the way: 'Strong pull-up' is called something that makes a circuitry look like having a source impedance of about 50...100 Ohm.

double checking of the pin value is not a hard job, but I wanted to know why it is absolutely necessary. this reasoning seems acceptable to me.

All this debouncing stuff is also needed, if you want to speed up response time of a keyboard: Assume, you want to be able to detect up to 5 subsequent key hits per second. Then you must sample the key faster than bouncing time lasts. Otherwise the user will get a feedback, which is not synchronous to his hits. When not polling the key very fast, the user will notice a certain delay to his action, a delay by the way, that is changing!! In certain cases, user might even think, that micro hasn't noticed his key hit, causing him to again hit the key.
A user, who is used to work with a fast responding keyboard, would find a slow responding one to be really nasty...

Kai