> The open question is whether Travis would consider whether open science is related to open source, hence whether it's legitimate to use their free CPU muscles

> The open question is whether Travis would consider whether open science is related to open source, hence whether it's legitimate to use their free CPU muscles to do science. The way Travis sets it up, it's free to run tests on any public repo. So OP's mechanism is a neat way to do repeatable (science) experiments, but I feel Travis would be in the right to say "hmm, no, you can't eat up tons of compute for free that isn't in the service of building software". That said, 45 minutes of compute on an n1-standard-1 is just 3.75 cents. So this only becomes a real pain if lots of people do it and start triggering say 100 builds a month ($5 is my mental threshold).