This may be normal behaviour - I'll try to explain why this could be the case.

NCrunch replicates build logic across servers in the grid. This means that if you are running a grid with 2 remote servers and local processing turned on, NCrunch will build your solution 3 times - once on each machine. This was a design decision made to try and simplify compatibility across the grid and to make it easier to operate the distributed processing with local processing disabled.

A side-effect of this is that the solution will need to be built on a remote server every time a connection has been established to this server. This is because the workspaces and session-related information on a remote server is cleared when the connection is lost, and needs to be recreated when the connection has been re-established.

So if you have a remote server that is at the end of an unreliable connection (for example, if you're using a laptop with a weak WiFi signal), and the connection is intermittently lost, NCrunch will automatically reconnect, then resync with the remote node and re-run the builds on this node.

Roughly how often do you notice this happening? Does it seem to correlate with disconnections from the nodes?

I've heard some reports of build/test spinners getting stuck perpetually when using the distributed processing, although to date no one has managed to pin down a way to reproduce it. In theory, the engine should drop all knowledge of remote processing when the connection to a remote machine has been cut ... though I wonder if there may be a hole in here somewhere. If you have this happen to you again, can you check for me whether there are tasks still processing inside the Processing Queue window? This will help to establish whether the issue is simply cosmetic or if it's a symptom of some tracking going haywire inside the engine.

Thanks for taking the time to give the distributed processing a try. I'm particularly excited about this feature and would like to ensure it is as stable as possible, so do post back if you have any further issues.

Cheers,

Remco