why

Because humans makes mistakes and learn from them. Its possible to have 0 errors on first run but probability of that gets low with every key you have pressed. It can be calculated like:

initialProb * (0.5 ^ keyPressCount)

Where initial probability is developers experience in domain.

It just depends on your assumptions as to what's an error!

If we can all agree that math is weird, and maybe 20+5 = -4195482287 sometimes and that core dumps are just the program telling you how it successful it was, then everything works! ....

*Cries in corner*