I discussed this in person with @timmaxw and @Tryneus -- this would be a spectacular addition to the reql admin release, and would take care of a number of related issues:

• long running query log (and ability to kill)
• much needed visibility into internal system processes

I propose to create a system table jobs, and include any tasks that might make significant impact on the performance of the system. Off the top of my head:

• user queries
• any time GC spins up
• secondary index post-construction
• replication

The user ought to be able to delete documents from the table to kill jobs (we may not let them delete everything; but certainly user queries and index post-construction).

I'm purposely omitting the structure of the document from the spec. I think the implementer should take a pass at it first and let us play with a prototype. Once we do, we can tune the API to make it more user-friendly (like we did with other parts of the reql admin).