Conventional theories of superconductivity assume that the slow collective modes mediating electron pairing are overdamped — their dynamics can be replaced by a single timescale, and the details of how many slow modes exist at each frequency don't matter. This works for conventional superconductors because the mediating phonons are fast enough that the approximation holds.
Chae (2026) shows that in correlated metals with many slow collective modes — strange metals, cuprates, heavy fermions — the approximation discards the essential physics. The time-scale density of states (TDOS) of relaxation modes controls the infrared dynamics. When the TDOS is rich enough, the system develops a 1/t memory kernel: the response at time t depends on the entire prior history, with no characteristic decay scale. This memory isn't exotic. It's what you get when you count the relaxation modes properly instead of replacing them with a single effective damping rate.
The consequence is dramatic: marginal superconducting instabilities in standard BCS and Eliashberg theory become robust algebraic transitions. The critical temperature scales linearly with the infrared spectral weight of slow modes. Superconducting domes and Uemura scaling — features that required ad hoc explanation in conventional frameworks — emerge naturally. The same slow-mode reservoir explains the anomalous normal-state properties (linear-T resistivity, long-time correlations) that have puzzled the field for decades.
The general principle: when a system's behavior resists explanation within a standard framework, the obstacle may not be missing physics but missing arithmetic. The extraordinary result was always latent in the ordinary calculation, performed without the approximation that seemed harmless. Counting carefully what you assumed could be averaged is not a minor correction. It is the mechanism.