oof2 users are forcing me to take down execution hosts because they use up all the memory available, and on top of that use a large portion of the huge swap space on disk. Performance suffers for all tool users because things have to be swapped out. I keep playing whack-a-mole with execution hosts, taking them down and back up when their sessions end.

I could lower the maximum memory that all tools are allowed to use in a container. However, that would prevent some simulations to be made at all. I could lower the maximum number of sessions that people can use simultaneously, but that would decrease the usefulness of the hub. I could shorten the timeout for that tool to less than a day, but that would be surprising behavior and would make it harder to view the results of long simulations. It would really be more appropriate to limit the number of simultaneous instances of resource-intensive tools, and control that separately from the other ones.