Evergreen task have recently started failing against the latest server. The cause has been traced to mongocryptd socket timeouts during test runs against replica sets and sharded clusters. While the root cause is as yet unknown, two simple changes have been found that will allow the tests to pass, and possibly improve reliability for users as well:

• Configure mongocryptd to log to /dev/null
• Configure socket connect and read timeouts for connections to mongocryptd