Comment 1 for bug 346017

Yes, I was already planning to fix this. We don't want to just ignore old ppi values, because there's some page sampling involved in guessing the ppi, and in case of a conflict the old one might well be correct, with the new one simply having missed the pages that require a higher ppi.

In general, the max value is usually the right one, but not necessarily, as the pdf may have been replaced between runs. So I was thinking when there's a conflict I'd do a "best of 3" or some such.