The question as to where the middlegame ends and the endgame begins does not have an exact answer. Sometimes the exchange of Qs is considered as the division, but this can't always be right because if it were, there would be no such thing as Queen endings. On the other hand, one sometimes runs into Queen-less middlegames.
A practical, although not an absolutely clear, indication that an ending has been reached is when the Kings begin to take an active part in the game. The question of the K's safety is no longer urgent and instead of being weak, the King becomes a strong piece, in many cases even the most powerful piece. The endgame is therefore characterized by (1) little danger for the King and (2) few pieces on the board.
One of the most frequent endgame tactics involves the creation of a passed Pawn. The Passed Pawn tactics involved are:
Forcing tactics for the purpose of obtaining a passed Pawn.
Advancing tactics the purpose of advancing a passed Pawn.
Promotion tactics which allows the promotion of a Pawn.
1) Forcing tactic example: Mieses vs. Wolf, Carlsbad, 1907. Mieses' 21st move is an example of this type because black can't capture both the c and d-Pawns at the same time and white's 22nd move will be either 22.dxc7 or 22.cxb6. As a result of Mieses' 21.d6 he obtained an outside passed P (the a-Pawn) which was stronger than black's passed center P (the d-Pawn). He also had the advantage of a B against a N.
2) Advancing tactic example: Euwe vs. Bogoljubow, Zurich, 1934. In that game, after black's 51st move Euwe had a K-side P-majority, but Bogoljubow's pieces were better placed, especially the Ks. Euwe's K was cut off while Bogoljubow's was attacking. Euwe's brilliant solution was to sacrifice two Ps in order to push his passed P one square which diminished the activity of Bogoljubow's R and allowed his own K to get into action.
3) Promotion tactic example: Johner vs. Euwe, also from Zurich, 1934. Johner sacrificed a P at move 41 in order to queen his f-Pawn, forcing Euwe to give up his B for the new Q. But, Johner had overlooked Euwe's brilliant little 45...e5 which prevents his B from stopping or capturing the h-Pawn.
Today's game between Lasker and Lisitsin, from Moscow, 1935, features an advancing tactic example. According to Euwe, Lasker's brilliant 46.d5!! saves the game. Using a couple of engines and running a bunch of Shootouts seems to indicate that, while the move 46.d5 is the clearest way to draw, white's position was not in the dire straights that Euwe claimed.
Personally, I think Euwe probably knew that, but he was using the game as an example of the "advancing tactic" in the endgame to illustrate his point. As a result, he avoided a thorough analysis. Why? Alex Yermolinsky wrote that general advice works for beginners but with more advanced players he confessed that sometimes GMs adjust the truth a little just to make things easier to understand. In Euwe's analysis he was teaching an idea and was not too worried about a couple of niggling little details. Comments?