Here is my use case:

• I have integration tests (in Java + with protractor) that are costly to execute (in term of execution time)
• when earlier hooks detect issues with the code (e.g. eslint checks), I don't want to execute my integration tests until I've fixed those errors

Btw, I don't ask to change the current behaviour, but could you develop on the reasons behind the design decision to run all tests, indifferently of their outcome ? What are the benefits ?