It would be cool to have a number of additional options for builtin health checks. I see the following being useful:

1. TCP connection open/close
2. HTTP (HEAD?) client access
3. Unix domain socket connect
4. File presence (e.g. a scratch file)
5. File "newness" (i.e. a file must have been updated within last n seconds)
6. Process presence (either by PID, or by process table search)

There are also some that are possibly extensions for others:

1. SQL query
2. Mangos/nanomsg req/rep
3. Mangos/nanomsg survey
4. Mangos/nanomsg pub/sub monitoring
5. ØMQ versions of above
6. Zookeeper?
7. object store access?

Also, we might recommend creation of a library & command to allow new applications to create their own watch dog facility that is light weight and more directly integrated. One can imagine, for example, a situation where an agent polls the REST API (using an app key I suppose) to keep a watchdog timer from firing.