I don't have a basis for how long this might take. As the author mentions "All bugs become shallow once they can be reproduced.", but only after spending probably the largest amount of time waiting for new incident reports to come in, and then analyzing the reports (e.g. to determine most incidents occurred on the last day of the month), and hours staring at application and kernel code. It's very impressive, but certainly the largest amount of time in the 10 month duration was not actually debugging. The "moment of extraordinary clarity" probably sprung out of years of experience.