Notes/Monitoring and ops aids

Monitoring

 * record ongoing rates for:
 * posting
 * registration
 * subscriptions

Maybe break down by:
 * total
 * per-user
 * ip/ip range?
 * user-agent/source?

notice_inbox table

 * really big
 * hard to keep under control
 * consider row per user instead of row per notice/user
 * evan + xopher chatting about rearch

email sending

 * currently done synchronously on notice posting
 * if mailer is borked up this leads to slow response or 500 error back to user
 * recommend moving to a nicer queue so the SMTP business can be deal with later
 * A local MTA on the app servers which can handle queuing and final sending offsite should resolve this without internal changes

xmpp

 * I get the impression ejabberd goes down a lot; we may need to find better ways to run or queue things