This reminds me of a paper from google, where they try to determine how risky a change to a file is, according to its commit history. Given you use semantic com

This reminds me of a paper from google, where they try to determine how risky a change to a file is, according to its commit history. Given you use semantic commit messages, you can just look on the number of fix() commits and infer a risk level. Later, depending on the calculated risk, you may run a different set of tests.