We don't need the lock to do that and it prevents a possible deadlock when
T1 holds the lock and tries to run an SQL query to resolve collection cache
policy, while another lock is holding the SQL lock and is waiting for the
CollectionScheduler lock to be released.