Looking for drains is too expensive as a regular operation. In production it may take up to two minutes just to run all the checks and Redis is what makes it slow. Tests have shown that no drain currently on a production node appears to be orphaned in either ETS or Redis, but always both. As such, we drop this lookup.
Channel deletion would delete only the ETS table entries for each drain, leaving the door open for new nodes and NSYNC'ing to bring back dead drains (that were not killed on existing nodes). This patch makes it so that deleting a channel, which calls logplex_drain:delete_by_channel/1 actually removes said drain via redis. nsync callbacks will then clean up the rest. Conflicts: src/logplex_drain.erl src/logplex_mon_serv.erl