In last week's post, TCEC Season 9 Superfinal Followup, I listed three opening variations which were won by White in both of the games where they were forced on the engine protagonists, Stockfish and Houdini. For two of those variations, the diagram below shows the position and evaluation one full move after the forced variation ended, i.e. after both engines were able to apply their own evaluation algorithms to the given position.
For example, in game 65, after 9...Be7, Stockfish playing White showed 'Eval = 0.54', Houdini playing Black showed 'Eval = 0.60'. I could also have chosen game 66, Houdini playing White and Stockfish Black, because the picture would be similar. In both positions shown here, from games 65 and 71, the engines agree that White has an advantage of about a half-Pawn.
In the top diagram, a 1.d4 Slav System, White has a better center, a space advantage, and perhaps a slight lead in development. In both games played with the variation, White eventually won a Queenside Pawn, which was sufficient to win the game.
In the bottom diagram, a 1.e4 French Defense, White has a space advantage, the better Bishop, and threatens to place a strong Knight on d4. In both games, White advanced the Kingside Pawns to create a blocked position where the Black King was immobile and passive, then played on the Queenside using the White King as an extra piece, eventually breaking through.
In the 'Superfinal Followup' post I called these 'busted(?) openings'. According to a calculation I did a few years ago, A Pawn Equals 200 Rating Points (February 2013), when one side starts with an advantage of 0.50, that's equivalent to an advantage of 100 rating points, which gives a 64% chance of winning. It's hard to overcome those sorts of odds.