Why are you trying so hard to prove that buffering is not a possibility? Maybe you should be spending your efforts on finding alternate causes if you think it's not buffering.

I gave the minimum lower bound.

If the buffer fills up once, there will be X bytes outputed. If the buffer fills out twice, there will be 2X bytes outputed. If the buffer fills out three times, there will be 3X bytes outputed. etc. The only time the minimium lower bound on the numer of bytes actually printed matters if it's lower than a reasonable X. 512, 1024, 4096 bytes are all reasonable Xs, so that's not the case here.

If you wanted to show weird buffering, you'd have to show the exact number of bytes printed is not a multiple of a reasonable X. The minimium lower bound on the numer of bytes actually printed is totally useless to this end.

How do you know the number of bytes that didn't print

I don't. But I didn't make claims of weird buffering.

In the absence of any other information, what else is there to assume?

Who says you should make an assumption? And if there's truly no other information, you could always assume the opposite.

...or unexpected data, such as \r or \b characters.

That would give the appearance of *starting* halfway, not *stopping* halfway.