arXiv2023
Extreme mass ratio inspirals (EMRIs) occur when stellar-mass compact objects begin a gravitational wave (GW) driven inspiral into massive black holes. EMRI waveforms can precisely map the surrounding spacetime, making them a key target for future space-based GW interferometers such as {\it LISA}, but their event rates and parameters are massively uncertain. One of the largest uncertainties is the ratio of true EMRIs (which spend at least thousands of orbits in the {\it LISA} band) and direct plunges, which are in-band for at most a handful of orbits and are not detectable in practice. In this paper, we show that the traditional dichotomy between EMRIs and plunges -- EMRIs originate from small semimajor axes, plunges from large -- does not hold for intermediate-mass black holes with masses $M_\bullet \lesssim 10^5 M_\odot$. In this low-mass regime, a plunge always has an $\mathcal{O}(1)$ probability of failing and transitioning into a novel ``cliffhanger'' EMRI. Cliffhanger EMRIs are more easily produced for larger stellar-mass compact objects, and are less likely for smaller ones. This new EMRI production channel can dominate volumetric EMRI rates $\dot{n}_{\rm EMRI}$ if intermediate-mass black holes are common in dwarf galactic nuclei, potentially increasing $\dot{n}_{\rm EMRI}$ by an order of magnitude.