In fusion, they can actually pinpoint exactly why it did not pan out as expected since 1953.
For fusion, you need to achieve a high temperature. So "thermal insulation" of the centre of the plasma from the edge of the plasma is really important. The nice thing is, that since charged particles move along magnetic field lines, and magnetic field lines never cross each other, the thermal insulation of plasma made from charged particles is enormous. That would imply it is really easy to achieve really high temperatures.
However, in 1953 they overestimated the enormous "thermal insulation" at higher temperatures by 10 orders of magnitude. That is a ratio of 1e10 between the expected and the actual value. It is still enormous, mind you, but not that enormous.
If that estimation had been correct, we would have had fusion by 1973. However, in that period it gradually became clear that fusion that way was not going to happen as people found lower and lower thermal insulations in their plasmas.
The same thing might have happened with semiconductors. What if the thermal noise were off by 10 orders of magnitude at smaller scales, compared to what was expected in 1953? Computing would not be where it is today, although it would definitely still be a good possibility if we were only pushing harder in circumventing the noise.