Well, should the problem not be related to the WinBoard version, but simply to the emergence of an engine version that was not around yet when you were using WinBoard 4.7.0, changing back to the latter will not solve it.
but it crashed already twice, after only a couple of dozens of games, and once it could not start one of the engines.
When you say 'crashed', what exactly do you mean? Does it exit spontaneously? (And if so, with or without error popup?) Or does it just hang? And if the tourney hangs, does WinBoard display a message in the PV field above the board, and does it still respond to the window close button?
When you play a multi-cycle tourney WinBoard suggests 'Sync after cycle' by default, so unless you explicitly unticked that when setting up the tourney, there is the possibility that at some point WinBoard will refuse to start a game, but in stead will just wait displaying 'waiting for other games'. This is intended behavior when there are still games of the previous cycle being played by other WinBoard instances. But sometimes this happens erroneously, as a result of a game that is not currently being played being marked in the tourney file as 'in progress'. This should normally not happen, because when WinBoard stops playing a game (whether it is through the game finishing normally, being aborted, or WinBoard exiting on user command), it should remove the marker from the tourney file that says the game is in progress. (But if the game for instance aborted because of a power fail or forceful killing of winboard.exe from the task manager, it would not have had the opportunity to do that.)
The way to check this if you experience trouble in the progress of a tourney is to have a look at the tourney file
when WinBoard is not running. The -results string (last line of the file) should then NOT contain any *. If it does, there is trouble for sure, as this marks the game as 'in progress', so that no WinBoard you start anew would ever pick that game. Replacing the * by a space with a text editor should repair this condition. (It is what WinBoard should have done when the game finished without a result.)
It could be that there still exist particular ways for a game to crash (perhaps when loading a new engine fails?) that would prevent WinBoard from removing the * marker for it in the .trn file, and that the engine you suspect knows how to tickle this bug. If you have trouble again, please send me the .trn file (after closing all WinBoard instances).