This is all getting way academic now, but if your estimate of those that use dbmopen is anything like correct, that's 1 in 100.

Whereas if my guesses for $^M are anything like correct, we are talking 1 in 100,000 to 1 in 100,000,000.

Even if every programmer added a $^M reserve to every script they ever wrote, your still only looking at 1 in 10,000 to 1 in 10,000,000.

Another difference is that dbmopen has been superceded. Before tie and all modules that use it became available, I think dbmopen was probably a lot more used. If you count the number of programs that use some form of tie, or tie-based module, your 1% is probably an under-estimate. Tie::File alone could probably manage 1%.

Nothing has superceded $^M.