Just a cautionary note - you are describing the future in the present tense.

"In three months"...there's a chance that a problem will be found by the people doing formal verification, but otherwise there don't seem to be any potential roadblocks. Client implementation is very simple, partly because most of the protocol is implemented by a smart contract, which is already done. Ethereum does a hard-fork upgrade a couple times a year.

51% attacks on the block proposal mechanism could prevent the PoS commitments from ever becoming finalized (i.e. violate liveness), though.

True, that's a risk it will still share with pure PoW. But at least you don't have to worry about doublespends after your transaction has finalized.

Full Casper may have stronger liveness guarantees eventually, I'm not sure. At a minimum it's easier to manually intervene to get the network going again. (You could also do that in PoW by changing the hash algorithm, but you can probably only pull it off once, migrating from ASIC to general purpose hardware.)