That paper is pretty old. There is a recent paper from a couple of months ago by the same author (Craig Gidney from Google Quantum AI) claiming that you could break RSA-2048 with around a million noisy qubits in about a week.
I
can't say for sure whether this approach can be applied to
ECDSA; I have seen claims before that it has less quantum resistance than RSA-2048, but I'm unsure if this is still considered to be the case. And while these papers are of course largely theoretical in nature
since nothing close to the required amount of qubits exists at this
point, I haven't seen anyone refute these claim at this point. These is still no hard evidence I'm aware of that a quantum computer capable of breaking ECDSA is inevitable, but given the rate of development, there could be some cause of concern.
Getting post-quantum addresses designed, implemented and activated by 2030 in accordance with the recommendations in this paper seems prudent to me, if this is at all possible. Deactivating inactive pre-quantum UTXOs with exposed public keys by 2035 should certainly be considered. But I still don't feel like deactivating pre-quantum UTXOs without exposed public keys in general is warranted, at least until a quantum computer capable of breaking public keys in the short time between they are broadcast and included in a block is known to exist - and even then, only if some scheme could be devised that still allows spending them using some additional cryptographic proof of ownership, ZKP or otherwise.
--
Best,
ArmchairCryptologist