When I'm using dotnet-trace, case normalization is applied inconsistently, and it appears we end up worse off than if there was no case normalization. The format doesn't matter - nettrace or speedscope both have the inconsistencies

eg dotnet-trace collect --profile cpu-sampling --format Speedscope --mydll.dll

Regression?

Probably not from my cursory review of the code

Other information

Practical effect of this is that speedscope viewers cannot group calls together:

ablaze-20260508-102713.speedscope.json