sue.loverso@mongodb.com is often able to reproduce format failures that other engineers can't repro. One thing Sue does differently is that she has her own scripts/method for running format.

When using the standard format.sh script to run N concurrent format jobs, it tries to keep N jobs running. So as soon as one exits, it starts a new one.

In contrast, as I understand it Sue runs N concurrent format jobs and waits for them all to finish, then starts another N. A consequence of this approach is that the load on the system varies more, changing timing within each format job. This may repro issues that just don't trigger under the load of "N jobs running at all times."

We should add an option to format.sh to test in this manner.

NOTE: I'm not suggesting/requiring that the option literally use the flag --sue.