Each platform has some crashes that are not real user crashes, sometimes as intended, because the teams want to gather data.

For instance, Clank uploads strictMode violations to crash, which go through analysis and eventually surface to crash front end, Fracas dashboard, and in the end sheriff queue after TEs file bugs. Then the sheriffs triage them.

It would be nice if the sheriffs don't need to triage them, since they are not real user crashes.

Potential solutions:

1) Don't upload them. They are not real user crashes, so it makes sense not to upload them to crash, but rather use an UMA metric instead.

Yet not uploading them may prevent more information from being gathered, since they may be used for more ambiguous cases that can benefit from full crash-like reports. If the pseudo crashes trigger the crash logic, then it is also more convenient for the teams to just upload them.

2) Have sheriffs not triage them. This is easy because each team has only a smaller team of sheriffs who are familiar with the platform specific issues.

However, this does not solve the problem that they get assigned to people after bugs are filed. A lot of attention and hours and compute resources have been used at this point.

3) Similarly, having TEs not to look at them, or having Fracas filtering them out requires the intermediate parties to maintain the list of hard-coded things to filter out.

It is not particularly maintainable, since it would be the teams that know which signatures/types of things to filter out and they may change from release to release as new features are implemented.

4) Upload them with a bit that indicates if they are pseudo crashes. So that the teams get their full crash reports, but the downstream analysis and tools can choose to deal with them differently without having to maintain a list.

This does, however, require changes on the whole pipeline of bug triaging and sheriffing, hopefully only as an 1-time overhead.

5) Have crash backend or Magic Signature analysis filter them out. Magic signature code already has a list of magic signatures.