I fully agree with you.
A disclaimer though, heh.
I truly believe, overall, that tactics trainer (and the like) is very much like endgame compositions.
Entertainment but educational, though if one is a pragmatic player, hardly useful.
An example. My good friend (rated unofficially @1705 USCF-he was officially rated at 1657 during the time of my anecdote) played out a position from Silman's Endgame course with me. At the end, he asked me just how I was able to solve it? (Obviously without having read the book, and just drinking my beer throughout).
"Dunno buddy. Wait. Actually I do know. I just calculated how many tempi I would need to get my King from g2, including pawn moves, to c5. Locking the pawns on the kingside and opposing your king (blocking him out of the center squares) and invading the "hen house" Silman's terminology. I was using Capa's Attrition principle.
Same thing, different generation!
Another recent example would be a YouTube video on the Carlsen-You game. Where the commentator asked the audience what was the right move to draw in the endgame position.
I silently said, "a5" because I was thinking b6 (via b4) is where I could invade the position. The host said that you might take some time figuring this out, try not to use your engines! Heh. Anyways, common sense in Chess is not really talked about these days. Fancy that.
I don't know much about computers but I was trying to explain some basic progamming that I did to my brother. It nothing related to any of this. He waived me away, saying "Brute force." I will never forget that. Anyway, a lot that passes for pattern recognition is brute force learning, inefficient and can be misleading. By the latter I mean that if you are relying on pattern recognition, you will look at a position thinking that you have seen this type of position before and look again and again until you finally convince yourself that there is nothing on. Obviously, if you had some way to telling you fairly quickly that there is nothing on, then that would be an advantage.