I'm pretty sure MAST only permits a limited amount of extra complexity using the existing op-codes.
MAST only increases privacy in specific circumstances; if a script has branches (e.g. when a threshold of signers is less than the total possible, 2 of 3 for instance), then only the branch that is satisfied ends up published on the blockchain when using MAST (e.g. if person 1 & 2 sign a transaction, person 3's pubkey isn't disclosed at all as a part of the transaction). With current scripting, all branches of a script are published when the script is spent from. This can make MAST transactions smaller than regular P2SH.
I don't think this will change LN privacy properties, it's already a private system in any case. I don't think it could make channel open/close transactions smaller, as they're always 2 of 2, but there's maybe some subtlety I've not picked up on.
There are no deployment implementations that I'm aware of, but any soft-fork is likely to get deployed using version bits (which Segwit was).