MCP server jobs are orphaned and not ending

[JasonTWalsh]

When any command is run through the mcp server, a QZDASOINIT job is spawned. When Claude is closed or the mcp server is reconnected, these jobs get orphaned on the server and are not getting closed properly.

I worked with Claude locally to come up with a solution to this to clean up the connection pools when the mcp server ends/restarts and have tested locally. The jobs appear to be closing as expected now so I plan on raising a PR for review.