He mentioned satoshi's coins as an example. He said that freezing his coins violates the principles of Bitcoin but if they didn't migrate, quantum computers might steal them eventually, dump them on the market, and harm Bitcoin as a whole. It feels like he is saying if we do something, it's bad but if we don't do anything it's also bad.
That's right. It is primarily a question of which action will lead to the least amount of harm. I believe leaving them as they are probably will do less harm, but nobody can know this for sure. If they are claiming otherwise, they are lying.
FYI source of that image is
https://chaincode.com/bitcoin-post-quantum.pdf page 18.
The least worst, in my opinion, is FALCON-512. Easier to verify (0.6x), and "only" 10x in size, in comparison with Schnorr. It will be 24x slower to sign it, but that's completely fine, IMO.
According to
https://falcon-sign.info/, FALCON-512 can perform 5948.1 signing per second on i5-8259U CPU. No one would notice 2ms to perform signing.
Signing speed is practically irrelevant compared to verification speed at least in terms of the traditional concerns relating to scaling the chain, I agree.