Our test harness produces a crash upon test failure. This intentional crash is indistinguishable from an unintentional crash which occurs after a test failure in TBPL without clicking through and examining the logs. Intentional crashes should be very quiet, and unintentional crashes should be very loud. cjones tells me this depends on bug 585187.
Not "depends on" in the sense of "we have to have bug 585187 to fix this"; we have another bug to suppress intentional crashes in the output altogether by cleaning them up. However, with bug 585187 we would be able to scrape all the crash stacks and either display the IntentionalCrash ones differently or filter them altogether.
These dependent bugs should eliminate all of the PROCESS-CRASH output for crashes we expect.
I think all the cases of intentional ignorable crashes won't cause a test failure now, and all other crashes will. The ignorable crashes shouldn't show up on TBPL as orange or in the blue suggestions box any more.