This is only the case if multiple keys are pressed, as stated. Decades of study have shown that if one waits until the key is released, what one gets is normally the correct key. If the user insists on pressing two keys at once, it doesn't matter anyway, since the result will be wrong in any case. If you're encoding a keyboard, lying on a flat surface and being "touch-typed," then the diodes may be necessary. However, if it's a keypad on the front panel of an instrument, the user is likely to be doing his keypad entry with one finger at a time and actually taking pains to ensure he's pressing the correct key.

The best protection agains inadvertent mistyped data is an "edit" routine which will allow the user to back up and correct the input before sending it to the calling process.

If you don't have any way of echoing the effect of the user's keypad entry, e.g. an LED indicator or an LC Display, such that he can verify/edit it, then, under ALL circumstances, you have a problem. Humans make mistakes. If you launch the nuclear missile or open the airlock on the space station on a blind keypad scan, you will get what you deserve.

RE