The problem is that, even on a single instrument (say a piano), if you start on some note (say C0), going up seven octaves should land you on exactly the same note as going up twelve perfect fifths (C7). The problem is that the precise tuning of that C7 is different depending on if you followed the perfect octaves or perfect fifths.
If you followed perfect octaves, which have a ratio of 2/1 (or just 2), C7 should have a tuning of precisely (2^7)·C0 = 128·C0. If you followed perfect fifths, which have a ratio of 3/2, C7 should have a tuning of precisely ((3/2)^12)·C0 = ~129.74·C0. The difference between these two is the Pythagorean comma.
Equal temperament solves this problem by keeping the octaves perfect (2:1), but compromising on everything else. It defines a half-step as the 12th power of 2, so a fifth ends up being a little narrower than a perfect 3:2 perfect fifth.
I wonder if there is any effort in electronic instruments to produce "better" frequencies on the fly? I'll explain:
On a piano hitting C+G plays something close to a perfect fifth.
An electronic instrument could move the G slightly to produce a perfect fifth when that key combination is played.
Edit: you could do it, but you'd better implement it on a feedback system that can re-tune in a hundred milliseconds or so.
What I wonder is mostly if that would sound any better.
Equal temperament solve that problem by normalizing the steps' absolute height. It's a trade-off between the instrument's perfect harmony, and the orchestra's overall one.