I strongly agree that the engine ("evaluation bar") causes kibitzers to be overly critical and claim blunders in non-obvious cases. (A good principle is "don't claim a blunder unless you can demonstrate a refutation on your own!") But in games 9 and 11, the GM commentators I watched (Giri/Polgar, Gajewski added in 11) described (
sans engine) why 27. c5 and 23. g3 were blunders, just before the moves appeared on the board.
Funny, I tuned in to game 11 right after 26. Kg1 and (even as a patzer) saw the forced win with 26...ef+ and the rook lift 27...Rd6. Of course I'd never have gotten there, as I'd have overlooked 22...Re4-f4 or any number of earlier moves. I was puzzled that Magnus missed the mate, but Giri pointed out that he probably saw a forced win after 26...Nf5 and didn't bother with other lines.