Everyone, I have compiled initial statistics based on 183 responses so far.
Everyone help out by voting!
I switched to a binomial-based distribution which is better for proportions (always exists between 0 and 1):
Plot of estimated failure rate during specific mileage intervals:
Plot of estimated failure rate before a specific mileage (perhaps more interesting to some of us):
(I didn't really propagate the uncertainty correctly, so the uncertainty is probably larger at higher mileages)
Looks like there may be a lot of failures between 60-75,000 miles, and a relative lull in failures between 75-100k miles (though still uncertain). This could mean that some pumps fail early and other batches are more robust. However, it could also be a function of the heterogeneous sample of turbo/non-turbo cars, different driving styles, etc...
Still, predicting a 50% chance of failure before 75,000 miles kind of sucks! (and ~65% before 100,000 miles)