In T340715 we could see that we got a performance regressions where more nodes needed to be re-calculated. However looking in the trace, it's hard to understand what's going on (we can only see that number of nodes increased, not why).
I asked in the performance Slack channel if someone new if there's more trace categories that can been turned and I got answer hat there are multiple disabled by default trace categories for devtools but not a specific one. Let me try that out and see what I can find.
- Test out trace categories
- If we find one, how much overhead does it add? If it's a lot, look into how Mozilla do it: they do one extra run where they get that extra data. That is implemented but we haven't used it on our side.
