Not a hardware person but it seems it wouldn't be that much of a difference?

Assuming bit flips happen at a fixed rate (cosmic rays whatnot), then wouldn't the likelihood only depend on the actual amount of memory used rather than the total capacity? Like, if a bit flips in unused RAM does it really matter... And used memory is used one way or the other so the flip would have happened anyways.

I think the bigger issue is outright RAM failures/errors are harder to diagnose. Like if I had a bad chip on one of the sticks, it would be harder to notice since the chance of hitting that error and crashing is much lower, so the system would seem to run fine most of the time. Doing memtest is rather annoyingly slow so I've only done it once after building the system to make sure none of the sticks were DOA.

On another note, for my NAS I do run ECC (4 x 8G DDR3) on a server board and over the course of about 7 years, I've had 1 stick fail in the form of consistently getting stuck bits (which are then corrected and reported) and have had about 3 or 4 spurious flips in total on the other sticks.