pull down to refresh
I don't think that any of these are blockers. After 30 years of engineering effort, we now actually have sub-threshold error rates in trapped ions, neutral atoms, and superconducting qubits. And we've seen no sign whatsoever of the errors being correlated in a way that would prevent fault-tolerance from working -- to their credit, the skeptics like Gil Kalai stuck their necks out and made a prediction, but their prediction turned out to be wrong. Meanwhile, just two weeks ago, there was a big advance reported in reducing the overhead for fault-tolerance, so that Bitcoin now looks vulnerable to systems with only ~30,000 physical qubits (see e.g. my blog for more details). What remains is to put all the pieces together and scale up -- which is engineering, not science. Like the situation with nuclear weapons circa ~1942.
The fault-tolerance threshold theorem is the basis of the claim that decoherence is a solved problem, at least in principle. But the theorem rests on three assumptions: sub-threshold error rates, sufficiently independent errors, and tractable overhead. Which is the real bottleneck for useful Shor, and which, if any, do you think might not yield to engineering effort?