@dampooo,
thank you for the explanation. I see your point, and would agree that Kb1 is the more practical move (preferrable even if Stockfish would have said it is evaluated 20cp lower).
In fact I also 'sensed' that my pawns on the queenside could end up awkward when I went for Qc2. But somehow I misjudged the situation and thought it was more accurate to bring my king into play. That strive of perfection and accuracy causing me trouble (but also what so often gives me good positions).
@AlbertoChueca, check your PM.
@lurarose, yes I spent one minute (out of my two remaining minutes) on the last move. Spending one more minute would not have helped. What I am trying to rectify is inconsistent/inreliable pattern recognition.
About that by now overanalyzed game: six months ago I would have sacrificed the exchange without thinking twice. This summer I lost an important OTB game because I gave a rook for knight outpost and a pawn, very similar to the one here. In that game it was the deciding mistake. If I instead would have undermined it, or tried to trade it off, the position would have been quite favourable for me. So in this online game I consciously did not consider a sac because I thought I could (and should) get rid of the outpost with other means.
Also, I do not play bullet. Tried it, dismissed it.
@risky-chess, thanks but the problem is not noticing threats - especially when I cannot spend several minutes on any move. Not how to play better bullet :-)
If anything, this is how to play better when I do not have 2-2.5h for 40 moves and then 1h per 20 moves. Even 90+30 is sometimes a bit too fast for me. :-(