Right now, vsmserver just launches all jobs (session verifications, etc) that it has received/scheduled right away, with no regard to any limits. We've seen that both low ports (bug 217) and file descriptors (bug 5442) has run out, leading to vsmserver crashes. It would be beneficial if we could limit the amount of jobs running concurrently to less than the number required to crash. With a queue, we could also make sure that we don't overwhelm any agent server with too many requests.