This isn't ideal behaviour, but I can understand why it is happening.

When you terminate the service, it forces the process to cease operating and shuts it down immediately. Because the service uses satellite processes (i.e. the task runners) which are performing other work, it cannot always clean up all resources on exit.

When the service restarts, it is designed to look for derelict workspaces left behind by the above situation and clean up them. Of course, in this case, it's now been restarted with a new workspace base path ... so it's looking in the wrong place.

I'll need to have a think about how this can be solved. Perhaps it can be done using a cleanup task from the configuration tool. For the time being, you may need to manually remove any artifacts if you change the workspace base path with the service already in operation.