They aren't though. No acoustic instrument has perfectly linear overtones. And even something like a flute, which probably comes closest, has an overtone system

They aren't though. No acoustic instrument has perfectly linear overtones. And even something like a flute, which probably comes closest, has an overtone system that varies with force and input excitation. The idea of the harmonic series is a useful approximation, not a physical truth. Fourier applies to completely static waveforms, and says nothing about how waveforms change over time. If you try use Fourier for dynamic waveforms - which is most of them - you run into all kinds of problems and tradeoffs. So the concept of whole number consonant ratios is an attempt to impose a mathematical ideal of perfection where it very much doesn't exist. Clearly music is related to the harmonic series, but it's related around it, not trapped inside it. The dissonances and deviations create colour, tension, and movement. This doesn't mean xenharmonic experiments aren't worth doing, and we should all just use 12TET and not think about anything else. But it does mean these are experiments away from real acoustics into computed perfection, not back towards a perfect ratio heaven that music was exiled from by 12TET.