How can I make this scenario work? ...

For various reasons, I may need to store many .rm files on a secondary server. So I can set the .ram files to point towards this secondary server ( if need be I could put the .ram files on the secondary server also ).

The question is: where must the Real audio stream engine reside - server 1 or server 2, in order for the files to play correctly?

The real server (engine) will be where your .rm files reside. And your .ram can be anywhere over http.

However if you want to use HTTP streaming, you don't need the real server and can put the files anywhere you want.