While I did mention that the sh glob is "cheaper" (which I now regret), please note that my post is about correctness and robustness, not performance.

I wasn't referring to searching subdirectories, only that specific use of ls. It serves absolutely no purpose whatsoever and can only result in mishandled filenames. In such a case, where ls is being used to retrieve the filenames in directories (non-recursively), always use file expansion instead.

While we're here, your find solution is also unable to properly handle any filenames with IFS characters (though it's not constrained by a command length limit if there are very many *.txt files). It could be made to handle spaces and tabs correctly, though not newlines, if instead of using the find in a command substitution in a for loop's list, you piped the output of find into a while read loop.

One last point: Please don't misinterpret my intention. I realize that most of the time technically-limited solutions are more than adequate; I use them all the time myself. I simply highlight the limitations so that anyone who is not aware of them will learn.

Regards,
Alister