I have trouble telling what's theory and what's my practical notes, but one thing immediately comes to mind.
Castling queenside can be nice in the Winawer, but it may be almost necessary in the MacCutcheon variation:
1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. Bg5 Bb4 5. e5 h6 6. Bd2 Bxc3 7. bxc3 Ne4 8. Qg4 g6 9. Bd3 Nxd2 10. Kxd2
You can see h4-h5 and Bxg6 are constant threats. Rh3 targets a rook lift, f6 is a hole for the queen or knight, and so forth.
Black can even play ...c4 (as in the Advance variation above) and target the weak a- or c2- pawns. White's pawns point towards the kingside, and an a5 push is hard without an exchange sacrifice. In this case, advancing pawns on the queenside makes it safer, since it locks the white c-pawns and shuts out the light-squared bishop.
If the c-file is up for grabs, the black king may be able to duck to b8 after castling and make it easier to trade off rooks. It's hard to open the a- and b- files, though a possible c5 or d6 outpost may be a nuisance. If the black king were on g8 after kingside castling, it would have ground to make up to get to the c-file or, in general, participate in the endgame. Your point about black's queenside pawns potentially moving and making weaknesses is a good one--you probably don't want to castle that way if you've moved them, the exception being a MacCutcheon-ish b6/Bb7/Nc6 formation, or where you've played ...a5 and White's played a4, and you control b4. Again, you want to make sure a knight can't slip into d6 or possibly c5.
But if pieces are traded off, or White's bishop on h3 is dangerous, ...Kd7 (hiding behind a wall of pawns) may also be quite good.
I think one side note is that, while castling queenside can be good, the central pawn bunker affords Black the option of delaying castling at times if there are other useful moves to make. This may not be theory, but it often happens if white blunders the d4-pawn in the opening. The e5 pawn can provide a kingside attack that is hard to defend in practical play. Black may want to push White's pieces away from d4 and castle to the wing White avoids.
Castling queenside may also be desirable in the Burn variation (...dxe4 instead of ...Bb4) since after Nxe4 Be7 Bxf6 gxf6 Black gets a hold on e5. There, castling kingside lets White castle queenside and have an attack with ready-made targets. Black should use the half-open g-file for some prospects, but staying in the center isn't an option as d5! may hurt.
If White has pointed pieces at the kingside and Black hasn't made a lot of queenside holes, queenside castling may be very strong indeed. It can, however, be vulnerable to a timely Ng5, threatening the f-pawn, if Black has not found the time to play ...f6.