Good Friday

I AM seeded 25th out of a final total of 44 entries in the Southend Open, which is being run as seven rounds over four days.
This morning I beat a 2137 after he allowed me me to trap his bishop on move 14. He played on but was unable to conjure a swindle.
This afternoon I reached the following position after 34 moves against another opponent rated quite a bit higher than I am.

White has just captured my queen on e5 and offered a draw in Tim Hebbes (2075) - Spanton (1914)

34...Nxe5 35.Re3 Rf5 36.Kg2?
The analysis engines Komodo9 and Stockfish10 strongly prefer 36.Nd2 or 36.h3, but note that going after the a4 pawn with 36.Nc3?? is a dreadful blunder because of 36...d4.
36...Kg6 37.Nd2 Ng4 38.Re2 Re5 39.Kf3
Or 39.Rxe5 Nxe5, when the black king and knight are better than their white counterparts, and Black has a protected passed pawn.
39...Rxe2 40.Kxe2 Kf5?
Over-sophistication. Black has big winning chances after the obvious 40...Nxh2. The engines give the main line as 41.b3 Ng4! 42.bxa4 Nf6, when it is hard to believe White can survive with all his pawns isolated and half of them doubled.
After the text, the game is equal.
41.h3 Nf6 42.Ke3 h5 43.Kd4 Ke6 44.b3 axb3 45.Nxb3 Ne4 46.Na5 Nxg3 47.Nxb7 Ne2+
This seems to draw, but maybe 47...Nf5+ was simpler.
48.Ke3 Nc3 49.Na5 Kd7 50.Nb3 Ke6?!
Again this may not be the most accurate. The engines prefer 50...Na4, eg 51.Kf4 Ke6 52.Kg5 Ke5 53.Kxh5, when Black is a pawn down but apparently draws with 53...d4.
51.Na5
The engines reckon White is slightly better after 51.Kd3 or, according to Komodo9, 51.Nd4+.
51...Kd7 ½–½
So, my Friday results were good, but I cannot say the same for my endgame technique.