Elected incumbents with support from supermajorities in legislatures have recently been primary drivers of democratic backsliding and subversion in some established democracies. Why and when do established democracies succumb to such threats? I argue that extraordinary events can give party leaderships of policy-seeking political parties in polarized societies an opportunity to select populist candidates who can help their party gain a qualified majority in the legislature required to dismantle horizontal checks and balances. Such democratic backsliding can enable that party's leadership to lock in their ideal policy. Yet, reduced checks and balances also enable populist incumbents to subvert democracy completely, thereby pushing the party leadership out of policy making processes. Impeachment through supermajorities in presidential systems and votes of no confidence through absolute majorities in parliamentary systems can constitute a credible threat against a populist incumbent's opportunism. But uncertainty about the fraction of the latter's loyal adherents in the legislature can undermine it. I develop a Markov game which shows that party leaderships may still have incentives to gamble on democratic subversion despite such uncertainty. Polarization between political parties reinforces leadership incentives to dismantle checks and balances. Heterogeneity of policy preferences within political parties reduces them. Most strikingly, parliamentary democracies are at higher risk to get dismantled than presidential ones. Once a presidential democracy has become a backslidden democracy, however, it is more vulnerable to full subversion than its parliamentary counterpart.