> Using the word copy in conjunction with C-NOT is slightly misleading as the copies do not behave independently.

That's what the word "copy" means. If you flip a coin and copy that bit, you will observe that those copies do not behave independently either. If you want independent bits, flip two coins.

Similarly, "erasing" a (qu)bit technically consists of performing a exchange operation between it and a known-zero bit. In typical electronic computers, this would generally involve diffusion-like exchanges between the voltage level in a memory capacitor (such as a FET gate) and that on the GND rail, which has a much greater effective number of bits and therefore will stay mostly zero, but eventually requires a thermodynamic expenditure of known-valued bits (aka negentropy) from some external source to maintain its voltage level / bit zeroness. (This is rather simplified; there's lots of other sources of known-zero and known-one bits getting depleted and replenished, and the exact accounting depends on how you interpret various physical states information-theoretically.)

> In typical electronic computers,

That should have had a "for example" in front.

Calling it a copy seems to me to violate QM's no cloning theorem, but maybe it's just that we have slightly different definitions of 'copy'.

The no cloning theorem is just the rigorous quantum mechanical version of the fact that if you flip a coin and write down the result, then copy what you just wrote down, you don't get two independent coin flips.