"Error correction that scales" is basically what the definition of a working quantum computer. That is less the breakthrough needed so much as what the breakthrough would lead to.
The relative lack of useful intermediate results raises major red flags for me, but it doesn't seem that physicists are all that bothered by it. Indeed, the high energy theorists have been at it for longer with a lot less to show for their work: At least we have a few toy quantum computers, while there is still nothing testable about string theory.
Honestly, I hate to say this, but I think it's going to take a major war for QC to have its watershed moment (a "Manhattan project"), and I think the odds are that it will actually work if it comes to that. However, the window is kind of closing on the usefulness of the technology as post-quantum encryption starts to get legs, and I don't really want another World War 2...
For AI it was the “attention is all you need” paper. I’ve been told that for QC it might be error correction that scales but I’m not knowledgeable enough to know if that’s true.
I expect that nuclear fusion will follow the same trajectory of being 50 years away until it’s suddenly 0 years away.
The reality for all these is that steady progress is constantly being made but isn’t visible outside experts in the field. Then it reaches a critical mass and you have the “chatgpt moment.”