I've used the "Waste an element" one for ages on microcontrollers where I don't want to deal with the overhead in an ISR.

Agreed.

The article author claims that the "don't waste an element" code is also more efficient, but that claim seems to be based on a hard-on about the post-increment operator, rather than any kind of dive into the cyclometric complexity, or even, y'know, just looking at the assembler output from the compiler.