firewalling - clarify please

I've tried to find a clear answer to this in the forums, but can't. This is sort of related to another thread about what vst(i)s to firewall. I understand the theory behind firewalling (separate plugin process, dedicated process per vst). What I don't understand is why you wouldn't want to keep each plugin in a dedicated process (or at least run all plugins in a separate process) all the time, whether known buggy or not. What are the performance hits (memory, cpu, compatibility, whatever) that you take by using this feature? What's the downside? Thank you for any clarification.

bump... 

The down side is processing. You can see for yourself by opening the same plugin natively and in a separate process and compare cpu use in the performance meter.

ok: * firewalled plugins are totally isolated from reaper (and from each other, if you select 'dedicated process per fx') * some plugins (e.g. synthedit-based plugs) get unstable if two of them are loaded at once in the same process. this is what 'dedicated process per fx' is for. * the more processes you have, the greater the general overhead on the system is (plugins in separate processes can't share resources). * there's also more communications overhead (reaper has to send/receive audio data over a dedicated pipe, instead of just sharing memory). this can be a BIG performance hit. does that help at all?

The biggest (and to me the only) downside of the entire plugin firewalling system is that it takes extra time for Reaper to communicate with all the isolated processes, so it affects low-latency performance negatively. The higher your buffer is, the less important that becomes. But try to avoid too much bridging in a low-latency situation.