Problem: dropping omni_httpd or database while it is being stopped #702

I may have observed an interesting case where while we're stopping omni_httpd, since it is down asynchronously, when we start dropping omni_httpd (or perhaps even the whole database), omni_httpd master worker errors out trying to query, and that leaves some http workers stranded, preventing the database from being dropped.

Proposed solution: may be we can have better exception handling to ensure we clean up the http workers in such case?

Workaround: make stop/start (optionally?) wait for the operation to complete? This would only be possible in the immediate mode.