The geological persistence of biotic assemblages and their reorganization or destruction by mass extinctions are key features of long-term macroevolutionary and macroecological patterns in the fossil record. These events affected biotic history disproportionately and left permanent imprints on global biodiversity. Here we hypothesize that the geological persistence and incumbency of paleocommunities and taxa aremaintained by patterns of biotic interactions that favour the ecological persistence and stable coexistence of interacting species. Equally complex communities produced by alternative macroevolutionary histories, and hence of different functional structure, may support less stable species coexistence, and are therefore less persistent. However, alternative communities with the same functional structure as a persistent paleocommunity, but variable clade richnesses, tend to be as or more stable than observed palecommunities, thus demonstrating that geological persistence is not the result of constrained patterns, or ecological locking. Numerically modeled food webs for seven tetrapod-dominated paleocommunities spanning thetraditionally-recognized Permian-Triassic boundary in the Karoo Basin of South Africa, show that incumbency before the Permian-Triassic mass extinction was maintained by a dynamically stable, community-level system of biotic interactions, thereby supporting the hypothesis. The system’s structure was lostthrough successive extinction pulses, and replaced initially by a rich but geologically ephemeral Early Triassic fauna, which itself was replaced by a novel Middle Triassic community with renewed incumbency.The loss of persistence and incumbency, therefore, did not result simply from the extinction of species; instead the largest declines were accompanied by the addition of new species to the system in the earliest aftermath of the event. We therefore further hypothesize that ecological reorganization and evolutionary innovation in the wake of mass extinctions play key roles in the destruction of highly stable, preexisting systems of biotic interaction. In the case of the Karoo Basin paleocommunities, we estimate that a return tostable interactions, and thus incumbency, was achieved in approximately 4-17 Ma.