Does Tessary fix the regression itself?

It proposes the fix, and a person ships it. Once RCA has named the cause, it opens a pull request against the connected repo with the change it believes corrects the problem, alongside the report: the hypotheses the investigation could support, the checks it ran and ruled out, and the files it points to. Nothing merges on its own, so the last call is always a review by someone on your team.

The investigation runs in an isolated session, working from the case’s trace evidence and the code to test explanations against what actually happened. Where nothing clears the bar, the report says so instead of guessing, and “no change located” is a valid outcome.

Two limits worth knowing. Without a connected repo there is no pull request, only the report. And a proposed fix is an argument from evidence, not a guarantee, which is why it arrives as a PR rather than a deploy.