DB connection pool leak in Listener/Content Editor

When I did kill -QUIT on the listener processID it returned several threads that were waiting for database connections (DatabasePool.get). See listener.txt.

When I ran "show processlist" it returned 76 connections, the majority were idle for over 5,000 seconds, most were for content_editor_google. See db-connections.txt.

This leads me to believe there is a DB connection leak in the listener code or the content editor code.

Unfortunately the listener log file was overwritten.

It seems to be related to either call or location extensions, because when QA ran a test to deploy 500 orders (without call/location extensions), they all eventually went through. But when QA was testing adding/editing/deleting call and location extensions, it locked up.