Assume that after elections two coalitions are formed. Let’s call them Player1 and Player2. Player1, although they can rule, wants Player2 to join them in forming a Government. Player2 refuses and aims for reelection where they believe that they will be in a better position. It is all a game of success and failure on the government to be formed:
So in the case where Player 2 believes that Player 1 cannot make it alone, they bet on their downfall in order to win the next elections whenever they are. And while Player 1 knows that they cannot make it, even with Player 2 on board, they push for their participation so as to make them irrelevant too in the next elections.
Any similarities to present day politics is purely coincidental.
Managing organized complexity 
Given the premise, this shouldn’t be described as a symmetrical game. That is, the time-constraint factor should be included in the matrix. Also, Player 1 is not Neutral to cooperation-success, while Player 2 does not perceive as loss the “All go down” option.
Indeed the matrix is a bit deceiving. A real game theorist (and not a mathematical tourist) could have made a better one.