AWS jobs can take far too long to clone repositories (hours)

From time to time, an AWS job running on production can take hours to clone a repository (especially nshm-conus) instead of under a minute. This happened back in January or February for a day or two before working as expected until March 13.

For example, on production:

• Job 9cbcc5fa8af878 took ~17 min to clone nshm-alaska, ~7 min to clone nshmp-lib, and ~19 min to clone nshmp-haz
• Job 0ea7c0556eefd0 ~2 hrs to clone nshm-conus and ~19 min to clone nshmp-haz

Communication with gitlab in a browser or on the command line in the office is not noticeably slowed.