This is one of the reasons why they never even brought a neutral into computer room panels back in the olden days.
You were just asking for trouble.
Many mainframes I used to work on were powered through motor-generators at higher frequency output, typically 400 Hz. That doesn't eliminate the issue, but at least it confines it.
If the source transformer was a wye/star secondary, there would still be issues at its neutral. If all your pulse current loads were connected L-L at 208V on a 208Y/120 system, you still have current peaks crossing the neutral junction in the transformer 360 times a second. Those junctions need to be beefed up.
Even then, harmonics could add 15% to the current on the delta connected line wires. And the I2R effect of narrow high current peaks is still there.
But at least you don't have to beef up an entire neutral run if you haven't connected to it. And power supplies are doing better these days, better approximating a sine wave in current drawn.
