I have resource share set to 10000 for my primary project (seti@home) and 0 for my secondary project (einstein@home). My intent is to have the hosts switch from seti to einstein each Tuesday when the seti servers are offline for 10 hours and my computers run out of seti work units. This seemed to work pretty well except some of the computers seem to start doing einstein work randomly even when they have plenty of seti work units available. Is there a way to use the resource share allocation to have einstein run only when the seti workunits run out? Thanks.

That should do it, did you check the logs to see if perhaps SETI wasn't available when the host asked for new tasks?

Unfortunately, when a 0% project requests work, it seems to get enough to fill your cache, although I think we would prefer it to get 'just enough' to get by.