The metrics query on sandbox will return incorrect resultt if I specify the start and end time in the request after some time has passed(in my experience, 8 - 10 hours), even though there is metrics emit during (end - start) time period. This happens on sandbox. I am not able to reproduce it on cluster. If I query not specifying the start and end time, the query will return the result.
This becomes important since now the profile detail page will show the metrics for the last 24 hours runs, node hours. These metrics will not be accurate if the query is not giving accurate result.
Another known thing is the number of log messages not shown up after some time on sandbox, but that is less important as long as we can show the actual log of the program run.