Some chunks reporting wrong running times

We are running experiment a28v that is a continuation of a legacy VHR experiment. Now we are using v3.13.0 and the experiment is running smoothly using vertical wrappers (this is a great success because these EC-Earth experiments using 5,000 processors are usually costly, slow, and unstable).

However, looking at the running times in the GUI, it seems that there are some inconsistencies. This is due to some inner jobs taking the first chunk's start time as their start time.

Probably the TOTAL_STATS files are not being updated correctly. This experiment is currently running so the logs are available in the exp directory.