You really need to read the manuals on Federation to understand it properly.
It work like this:
You have 2 servers S1 and S2, and a table T1 on S1.
S2 has a nickname F1 that is a federation link to T1, so queries against F1 on S2 are automatically routed to S1 to read T1, and the result is sent to S2 to return to the client. As far as S1 is concerned, S2 is just another client. So if T1 is dropped, S1 does nothing more. Then when F1 is accessed again on S2, it will fail because T1 no longer exists.