Using RSE, from what I can see, would require adding the RSE runtime to our build, which would add another 4-5 meg of requirements and 'bloat'.
Rob, stop calling more functionality bloat. And 4-5 meg in comparison to rest of WTP/Eclipse is lightweight!

AND if our AS adapter is modularized enough we wouldn't need it for everything...similar to you dont need TPTP installed even though we integrate with it.
Also, to have a 'server' to publish to, from what I can see, all you really need to be able to do is be able to make directories, copy folders into them, and delete files and folders. Using raw JSCH seems to allow me to do that very easily. But I wonder if maybe my use case list is too narrow and there's things I'd be missing out on.
remote start and stop.

And the usecases you mention seem to be exactly what RSE covers and freely available if you use the resource API ?
It seems most of the benefit of RSE is gained by using this remote connection to hold your projects, files, etc, almost as a remote workspace. According to a blog post, the benefits of RSE are:

RSE is a framework and toolkit in Eclipse Workbench, that allows you to connect and work with a variety of remote systems, including
I'm not sure any of these are requirements for a server to publish to.

Max, any thoughts?
It would be interesting if RSE allows you to mount remotely without exposing it directly in the project view...

I'll need to play a bit more with RSE to answer better.

btw. how do I get to play with your stuff ?