Fast blue optical transients broke the supernova taxonomy. AT2018cow, the prototype, rose to extreme luminosity in days rather than weeks, stayed stubbornly blue and hot, and kept radiating variable X-rays and luminous radio emission long after any radioactive-decay-powered transient should have faded, the signature of a compact central engine driving mildly relativistic shocks through very low-mass ejecta (Margutti et al. 2019; Coppejans et al. 2020). A handful of cousins, some in dwarf hosts, some off-nuclear, share the pattern.
Each standard channel fails a different requirement. Core-collapse supernovae cannot rise that fast or stay that blue without unphysically little ejecta around an engine that conventional collapse rarely produces; tidal disruption events by massive black holes struggle with the off-nuclear locations and dwarf hosts; intermediate-mass black hole disruptions require a population whose existence is itself unestablished; and magnetar-engine models need finely tuned cocoon geometries and jet orientations reproduced across the sample without overproducing ordinary events (Perley et al. 2019; Lyutikov and Toonen 2022). The class is small but the strain is structural: every explanation borrows an engine from outside the standard stellar-evolution inventory or tunes one inside it.
The standing is young and data-limited, with survey cadence the bottleneck: Rubin-era time-domain coverage will multiply the FBOT census and fix the rates, hosts, and environments that discriminate among engines.