While it is tough to completely solidify "potential keepers", I think some players might be a given. I've never done a keeper league, or an AL-NL-only league for that matter, but here are my thoughts.
To me, I'm not so sure fantasy managers should have to keep track of "the likelihood of a player to get traded to the other league." Yes, maybe they can get a sense of which players will be FA soon, who is unhappy in their current team, etc. But the likelihood that they get traded to the other league is almost the same that they get traded to the same league (Ok, maybe the chance is slightly less of staying in the same league because GM's won't want to trade top players to their competitors, but I don't think it's THAT much of a difference).
I think an interesting way to handle this would be to look into how MLB handles "sandwich picks" when dealing with teams losing FAs. Basically, as you suggested, it would involve assigning compensatory picks to teams who lose players to the other league. I don't know if the public has access to this, but maybe you can find out how MLB rates FAs (Don't they fall into different classes? Like a class "A" FA would warrant a sandwich pick after the 1st round, a class "B" FA would warrant a pick after the 5th round, etc.). If you can't find that info, maybe you could give it to one of the managers in your league as a little side project. Have them do some research about player values and assign them all a "class" in case they change league. Then, as a league, decide where to stick your compensatory picks in the next years' draft. I think this would eliminate managers claiming thet were going to keep someone, just to gain a draft pick. If all the compensatory picks were at the same time, you would see a lot of that. With the classes, they will only be able to draft someone of equivalent value, you know?
Sorry this is kinda long-winded, and it may even be too complex of a solution. Just an idea though, because I think mimic-ing how MLB does it is generally a good way to go.