> Do you have a way to revert to the old installation to check whether
> the checks fail again? It might be useful to track down exactly
> what happened. It seems wrong that a currently-installed version
> should have adverse effects on a just-built version's regression
> tests.
No :)