We should create a precise specification of what the Mesos source code is supposed to be implementing wrt. the life cycle of executors and tasks. And in addition, we should document why certain design decisions have been made one way or another, to provide guidance for future code changes.
With such a source code-independent specification, we could write unbiased regression and scale tests, which would be instrumental in maintaining high quality.
Furthermore, this would make the source code more amenable.
Why pick this particular area of the source code? Shouldn't more of Mesos have a thorough specification? Probably so. But we need to start somewhere and this area seems to be a good choice, given both its intricacy and its importance.