The conventional approach to conserving threatened biota is to identify drivers of decline, instigate actions to mitigate threatening processes, and monitor interventions to test their effectiveness and ensure target species recover. In Australia, predation by introduced predators is a threatening process for many native mammals. Here we report the results of a 15 year monitoring study in an iconic Australian reserve, Booderee National Park, where exotic Red Fox (Vulpes vulpes) populations have been controlled through an intensive poison baiting program since 2003. Unexpectedly, we documented the collapse of native mammal fauna during this period, including fully arboreal species that should be largely unaffected by fox predation – such as the nationally Vulnerable Greater Glider (Petauroides volans) and Common Ringtail Possum (Pseudocheirus peregrinus). We used path analysis to explore potential causes of these unexpected declines. We found no compelling evidence to support hypotheses that competition with increasing native species, native predator release, or increases in native herbivores underpinned mammal declines. Beyond the path analysis, data from other studies completed both inside Booderee National Park and outside (where intensive fox baiting does not occur yet depleted fauna species remain), allowed us to rule out several drivers of change. The temporal declines we documented for arboreal marsupials were not anticipated nor explained by any clear mechanism. We propose the use of experimentally-guided reintroductions and translocations to: (1) restore empty niches such as the currently vacant apex mammal predator niche, (2) reconstruct the now depleted arboreal marsupial guild, and (3) further test key hypotheses associated with mammal decline. We also suggest that given the potential for perverse outcomes following large-scale management interventions (even those where there is high confidence of success), wildlife managers should consider maintaining reference areas (where there is no management intervention). Finally, as the declines we documented were unexpected and rapid, there is a clear need to develop more sensitive early warning signals to alert conservation managers to impending problems, allowing them to alter management regimes before major declines occur.