I went and looked at my old performance exercises and that's essentially what they did in that. There was the build array by appending and prepending. You had the option to turn a reverse boolean on for the append to flip it and show the time difference between just the append and flip.
In that case, the reverse was negligible. You were looking at a "time" of 3-5 with, or without, the reverse. That would obviously change with the reverse, append, reverse even if only a little.
Even if Darren doesn't come here to offer clarification, his statement in the white paper linked up above was pretty clear. I'm not sure why he didn't build that into the compiler if it's faster. But, he's stated it's faster.
