I approve of being cautious with a guy who is going to have to pitch a lot deeper into the year than he ever has, pile up a very big innings jump year over year, has had shoulder inflammation, and when we have a decent lead. We joke some about stats here, but the Win Probability estimate at the time he was pulled was about 85% (3-0 lead on the road after 6 innings), while pitching him another inning of shutdown ball would only increase the WPA 4%. I'm not saying those numbers are definitive, but it is a good indication that the Nats were very likely to win that game, that pitching him a little longer wasn't going to help all that much. When you add in the bullpen had had a day off, it was not like you were stressing them.
Pitching from 87 pitches up to 100 - 105 pitches, assuming things go well, would add stress to the arm at point in the season where you want to conserve stress and had an opportunity to do so, even without the shoulder concern. If things did not go well, you were not going to let go over 110 pitches to let him clean up his mess. That's real pitcher abuse in a game where you have a lead in the division and are not facing ATL or a team you are in the playoff mix with. If things start to fail, then you need to bring in reliever with men on - a much more difficult "hi leverage" situation than starting the 7th with a 3 run lead and no one on.
This decision made all the sense in the world. Sounds like hindsight, but just because everything worked out consistent with the probabilities does not make the probabilities an invalid explanation of the strategy.