Better monitoring

In #14 we concluded that the playground could do with decent monitoring, so that we can enable the humans (human?) running the server to detect abuse that the bare-bones IP-based spam detection doesn't catch.

I'd prefer it if said monitoring doesn't bring in a boatload of additional software, though it may be unavoidable to bring in an additional daemon/server that aggregates events.

Start by reading issue #14 and the discussion here to understand the existing IP-based spam detection and the monitoring needs. Define an approach that lets server operators detect abuse while avoiding unnecessary additional software; done means those requirements are documented and the project has an agreed implementation direction.

Written by the indexing model from the issue text.