Software engineering must be a special case. Half of it is about handling uncertain states and preparing for unknown future uses.
Engineers need to build things with safety thresholds. Most devices will probably experience situations where it is accelerating at a greater rate than just that due to gravity. Engineers work in a messy world without controls and try their best to deal with non-ideal scenarios. They live in a world where a cow often cannot be idealized as a sphere. Because of this, the precise acceleration often doesn't matter, what matters is that the device works exactly within some range of operating environments. Only in cases of metrology do engineering devices need precise measures of external factors.
Meanwhile, scientists need precise measurements of environmental influences because they seek to eliminate them from experiments. otherwise, the signal cannot be distinguished from the noise. That is why the mass of an electron is known to an exaggerated precision, not because an engineer needs to build a device for it, but because (gross simplification) the precision is needed to be able to subtract electron related signals from the Higgs boson signal.
Have s look at an IC engine and the tolerances required for an engine to work.
Let alone say correctly designing nuclear rectors to handle two phase flow in its cooling loops.
You can't count on wafers being exactly at the same size, so you build your machinery in a way that precise sizes aren't important, only relative measures within a wafer is.
You can't count on alignment measurements, so you build your machines to use the only alignment invariant available, that is the size of a single wafer (and call it self-aligning).
You can't count on features having any specific size. They vary wildly, in proportions that in any other specialization of engineering would be disastrous. So, you just manufacture them, test after the fact, and throw away the parts that do not comply.
The entire field of IC manufacturing is about how to not need measurements and be resilient to errors.
Also, I don't know much about nuclear reactor design, but I'd be very surprised if they used measurements with several algarisms.
In this regard, the practice of software development (as opposed to the state of the art) is mostly distinguished by an aversion to seeking answers to these questions before developing a solution. Software engineering attempts to change this, but in practice, it is often pursued through ad-hoc rules-of-thumb of the sort that craftsmen use, rather than anything equivalent in rigor to physical engineering's foundation in the physical sciences.