The Fornax dwarf spheroidal carries a clock that has been telling the wrong time for fifty years. Five old globular clusters orbit the galaxy at radii of order a kiloparsec, including the massive cluster Fornax 3, and in a cuspy cold dark matter halo, dynamical friction should have dragged them spiraling into the center within a few gigayears, piling them into a nuclear star cluster (Tremaine 1976; Cole et al. 2012). The clusters are ancient; the nucleus does not exist; the clusters remain where they are.
The timing problem's standard solutions each concede something fundamental. A large constant-density core in Fornax's halo stalls dynamical friction, an effect called core stalling, and N-body work confirms it keeps the clusters aloft for a Hubble time (Read et al. 2006; Boldrini et al. 2020), but cuspy profiles are what cold dark matter simulations generically produce, so the cure is the core-cusp problem wearing a different hat. Tuned orbital initial conditions, placing all five clusters on wide orbits initially, survive only as fine-tuning. The problem's persistence has made Fornax a standing citation for cored halos, modified gravity, or non-standard dark matter.
The standing is one of the cleanest dynamical arguments in the small-scale sector, because dynamical friction is textbook physics: if the medium is there, the clusters must sink. They have not sunk.