Much of the SPT codebase is good at generating detailed logging, but there are a few specific issues that impair the process of reviewing the logs:
Too fine-grained reporting by the computation workers. These should perhaps be consolidated into rate-limited summaries, e.g. after a few minutes report the number of jobs taken off the queue, computed, etc.
A few mistakes in logging logic that create unnecessary warnings (specific example: the feature request / scheduler reporting that "denominator" samples are missing due to mixing up denominator and numerator).
Much of the SPT codebase is good at generating detailed logging, but there are a few specific issues that impair the process of reviewing the logs: