Probably already known since this is still experimental - for observable_source_asset jobs, there isn't an "Observe sources" equivalent button to run the job on demand. Once there's been at least one run, re-executing is one way to do it. It looks like the "observe" equivalent for "materialize" is missing in most places. That "observe sources" button from the assets group works fine in the mean time.
The overall direction with observable source assets is good I think. For now, I don't think I'll use it in replication tasks when there's a single source table and single target table, due to adding a layer of indirection (single-click materialize becomes "observe and wait for async auto-materialize, then navigate to see results").
I think I've got a form of global op concurrency control working well enough for my purposes with dask-distributed Semaphores handling locks for multiprocess executor. Added notes in the earlier thread.