Dynamical arrest — the freezing of a system's dynamics into a long-lived plateau where it stops relaxing toward equilibrium — is normally attributed to disorder. Spin glasses freeze because random impurities create frustration. Structural glasses arrest because random configurations create trapping basins. The disorder provides the landscape; the system gets stuck in it.
Kundu, Seth, Roy, Bhattacharjee, and Moessner (2602.23362) find dynamical arrest in a spin-3/2 ice system with no disorder at all. The system is perfectly ordered. After rapid cooling, spin autocorrelations reach a plateau that persists for exponentially long times — the system stops relaxing, not because it's trapped in random disorder, but because its own composite excitations (magnetic monopoles interacting with crystal-field excitations) create kinetically constrained pathways. To relax, the system would need activated processes — coordinated rearrangements that become exponentially unlikely at low temperature. The barriers are intrinsic to the clean dynamics, not imposed by randomness.
This is a disorder-free Coulomb spin liquid that behaves, dynamically, like a glass. The phenomenology matches: long-lived memory, slow relaxation, apparent freezing. But the mechanism is different. The conventional glass has extrinsic barriers (random landscape). The spin-3/2 ice has intrinsic barriers (composite excitation structure creating kinetic constraints). Same arrest, different architecture.
The general principle: when a phenomenon is associated with a mechanism (disorder causes arrest), finding the phenomenon without the mechanism forces a distinction between the outcome and the cause. Arrest requires barriers. Barriers can come from randomness or from structure. The randomness was never the explanation — it was one source of the barriers. Removing disorder and finding arrest anyway reveals that the explanatory weight was on the barriers, not on the disorder that usually provides them.