Is engine analysis actually useful? (Example from a game of mine)

In a recent game of mine my opponent played Bxd3, which is apparently quite a mistake according to Stockfish.

Stockfish's evaluation jumps from +2.5 to +5, so after this I'm like an entire rook ahead... But what's the best next move? That's actually not at all self-evident.

The move I did in the actual game, Bxf6, apparently also is quite a mistake because Stockfish's evaluation jumps from +5 to +0.6, so it's like losing an entire rook worth of positional advantage. But if that's not a good move, what is the best move?

Apparently it's Qf3. (There are other moves that are worth like +3, but this is the only one that's worth +5.) But how would have I known that this particular move is the best one?

The queen does attack the b7 pawn, but that pawn is almost inconsequential because of being behind a doubled pawn, and taking it isn't even part of Stockfish's main line. Apparently the reason why Qf3 is good is because the best continuation goes like this:

Evaluation is still at +5 here, ie. an entire rook's worth of positional advantage. Black is apparently losing some pawns here. However, there's no way I could have read this far ahead, especially given how many possible other answers my opponent could have given along the way, making the amount of branching exponential.

Also, even this particular position isn't so clear-cut why it's worth +5. After all, black can still do some tricky shenanigans by playing Re1+ and then harassing the bishop and the pawns on the queen side, so it's not clear at all how this is so much advantageous for white.

I can't see any other principle by which that Qf3 would have been somehow evidently superior. This just feels like one of those moves that only engines and really strong players can see, maybe. I don't think I have learned anything from analyzing this position with an engine.

So is engine analysis really all that useful? It doesn't teach any principles that would be beneficial.

One engine analysis line being hard to follow does not lead to engine analysis itself not being useful. Your case doesn't really hold up or make much sense.

What Catmaster0 said.  Don't just look at one engine line if you're trying to learn themes or principals.

How many ply did you analyze?  That will also influence what the computer is seeing.

My suggestion would be to look at the top moves that are fairly closely evaluated.  Maybe look at everything within half a pawn.  Play each of them for 3 or 4 moves.  Look at the positions after.  Which position looks best?

Qf3 is threatening to force black to double pawns on f3 after BxN, QxB and QxQ.

Here's one possible outcome...

I see a very weak pawn on b7.  Black is going to want to exchange bishops, otherwise this becomes a great advantage for white.  This looks much better for white than black.

With your scenario, it looks about equal.  And the computer evaluates it that way, too.

My engine gives Qf3 as +1.1 vs +1 for h3 or +.9 for Be4/Bg6 vs +.0.1 for Bxf6.

I would look at exactly how long you're letting Stockfish run for the +5 eval.  It doesn't make sense to me.

It seems like you're asking how to get better at chess. I don't think engine analysis is particularly useful for players of our level (we have similar daily ratings). An engine move and moves by good players are parts of sequences of moves and parts of plans. if you don't understand what their plan is then analysing that game isn't going to help you much. You can learn the value of their moves (in this case attacking an undefended piece, linking up the rooks etc) and assimilate that value into your existing chess knowledge, but expecting to become as good as an engine by seeing which moves they play isn't helpful in my humble opinion. Also don't get too hung up on the +5's and +3's and all that. You can be +10 but if your pieces are on poor squares your material advantage won't help you. Solid positional play will lead to small advantages which will sometimes turn into decisive advantages. Engines only really useful for the best of the best. At our level working on the fundamentals will eventually allow us to see better and better moves and so computer analysis will be of more use to us when we possess more skill.

It quickly jumps to about +5 and stays there no matter how long I let it run. For example I let it run for a while and got +5.62 at depth 50/82, 3.5 billion nodes calculated.