Over the last few days I've been flicking through a couple of old Russian endgame books by Averbakh et al. The books (which were published in English by Batsford) were part of a project to write an endgame manual for soviet chess players. It turned out the task was too big for one book (or one author) and so it is divided up into different types of endings.
One of the things I looked at was the accuracy of the answers in the book, seeing if they would stand up to modern tablebase analysis. I took a selection of interesting position and put them into the http://www.k4it.de/index.php?topic=egtb&lang=en tablebase page and checked the answers. So far I haven't found any significant errors (just the odd alternative move deep into the solution), and no instances of the wrong assessment being given.
One position I looked at was this study by Grigoriev from 1931 (White to play and draw). The answer involves the two kings dancing up and down the board, making sure that the capture of one pawn is met by a capture of the corresponding pawn. There are a number of 'only' moves needed to solve this position, and this is confirmed by the tablebase. I also fed the position into Stockfish, hoping to trick the computer engine, and for about 1.2 seconds it thought that Black was winning. The it hit a search depth of 31 ply and concluded the position was drawn
btw the bare bones solution is 1.Kg3 c5 2.Kf3 Kf1 3.Ke4 Kf2 4.Ke5 Kf3 5.Ke6 Kf4 6.Ke7 Ke5 7.Kd7 Kf4 8.Ke6 Kf3 9.Ke5 Kf2 10.Ke4 Kf1 11.Kf3 Ke1 12.Ke3 Kd1 13.Ke4 Kd2 14.Kd5 Kxd3 15.Kxc5 Ke4 16.Kd6 =