André wrote:
Plan C could be to do what testbot does and remember possible fails, maybe including the out of the blue ones.
Testbot still suffers from a lot of false positives, I think.
Yes, a build bot can do all sorts of workarounds, but we shouldn't forget about normal developers. It would be great if the average developer got in the habit of running 'make test' frequently and could rely on it not failing randomly.