Of all the answers here that provide stats, I wonder what percentage of those have actually measured the time versus just giving numbers that 'feel about right'

Of all the answers here that provide stats, I wonder what percentage of those have actually measured the time versus just giving numbers that 'feel about right'. If there is one thing I've learned about stats it is that 96.60% of people's gut feel is usually wrong ;-) That said, about the question asked. I can attest to the fact that in general thinking and writing good tests greatly reduces time spent debugging. Also, adding judicious and useful logging for the critical / edge cases helps a whole lot.