just another data point in favor of routing the cables under the bar tape to the stem area.

I've done both - exited the cables right below the bar curve to loop in front of the handlebars and then back to the down tube, and then next time I cabled, I tried just routing them under the tape all the way around to the stem.

Both worked perfectly. Sure, there's "theoretical" additional friction on the cables which "might" degrade shifting. But in my case, if it was there it was imperceptable.

FWIW, the housing length was very close to the same for both techniques, so that was not a factor at all. The only possible factor was that the cable makes slightly (slightly) sharper curves with the second technique, but again, it was meaningless.