We have several repositories sync’d and translating fine, but noticed the server hosting it was running at memory capacity. We bumped it from 16 to 32GB, but ran into the same problem. We have been restarting the translations daily as a workaround since it takes a while to max out, but were hoping you may have some insight to a permanent solution.
as a first step I would recommend checking if you are running the latest version os SubGit, we have implemented some resources consumption fixes in recent versions, so it would worth upgrading it to the latest 3.3.12 version.
Could you also advise is that an initial replication or the initial import has already finished and the repositories are now just being synchronized? If the initial replication has finished already, could it be possible to provide us with SubGit logs from an affected repository and also
jstack -l PID output for the SubGit daemon process for the same repository?