This is borderline absurd. It's the same reasoning that was applied to this post:

What are you using as the foundation of your decision?

1) Your opponents won't make mistakes
2) You shouldn't be looking for the most correct plays, but the ones that keep you out of trouble
3) You're counting on screwing up in order to knowingly reject a good play in favor of a bad one

This is the type of thinking that leads players to make passive plays when you they should be aggressively trying to establish an advantage. It's the same reasoning that leads to players who bury checkers out of fear of getting hit. It's the type of reasoning that leads players to "give up *substantially more* equity in the future." It's a compounding problem that affects much more than just this position.

I think it's bad to take this type of mental approach to the game, but I think it's especially bad to do it on a quiz problem. It doesn't matter to me whether we agree philosophically on this point. I very strongly believe you're causing more self-inflicted wounds than you realize and I'm willing to take a hard line with it.

These positions are actually very easy to play. All your moves are practically forced. Either you hit a late shot and win or you don't.

It is a great deal easier to play than the initial position.

One last comment about backgames:

Backgammon Boot Camp on page 176 at the bottom seems to imply that a bot might tend to underestimate the winning chances for the player who has the backgame, because "all the tough checker-play is on White's side". I think he's saying that there is an imbalance of errors in the rollouts because white's proper play may require long-term planning that bots may not be able to do effectively, whereas all of black's plays are short-term considerations that bots can do effectively (ideas such as minimizing shots in the next one or two rolls).

If you accept this assumption, then in some sense the value of the backgame according to the bots already has some accounting for errors of play and that an expert should actually be about to outplay the equity values that the bot would give.

The book has a copyright date of 2004 and I don't know whether bots have improved significantly in the past 5 years in this area of play to make the observation obsolete.

The bots themselves haven't changed at all since around 2002-2004 (NN, not features), though I think some of the theory & practice has.

Both GNU & Snowie should play this fine. They both can build boards and hit shots, which is pretty much all the expert and/or below-average open player is going to be doing.