Ask HN: How do you debug test failures in CI?

It seems like developers frequently run into situations where the only way to debug a test failure in Continuous Integration is by repeatedly pushing incremental changes until a smoking gun in the form of something like a log line is found.

How do people solve this? Should we accept it as the cost of doing business?