Open an issue to run the benchmarks.
GitHub Actions are not an isolated environment so statistical outliers can happen.
I use issues (labeled event) instead of workflow_dispatch as trigger because it allows to have easily only one issue for a PR (if you need to run the benchmark several times).
I use issues (labeled event) instead of issue comments to limit the number of useless triggers on all comments.
One approach can be to use workflow_dispatch as trigger and add the comments inside the targeted PR, this will work for PR, but the benchmark of all linters (not related to a PR) will not work.
The simple run outputs are limited because of the body limit of a comment (maximum is 65536 characters):
- to max 500 lines
--max-issues-per-linter 50--max-same-issues 10