In a standard SIR epidemic model, lifting interventions too late wastes resources but lifting them too early simply means the epidemic resumes from where it paused. When pathogen evolution is added to the model, this symmetry breaks. Lifting interventions too early can produce outcomes worse than never intervening at all. The mechanism: suppression slows transmission but does not slow mutation. The pathogen continues to explore its fitness landscape during the intervention period. If the intervention is lifted before the susceptible population has declined sufficiently, the evolved pathogen meets a population that is both still largely susceptible and now facing a more infectious variant than the one the intervention was designed to combat.
The researchers also identify an asymmetry among intervention types. Measures that shorten the infectious period (isolation, treatment) reduce transmission but do not suppress viral evolution -- the pathogen still accumulates the same number of mutations per infection. Measures that reduce transmission rate (masking, distancing) accomplish both: fewer infections mean fewer replication cycles, which mean fewer evolutionary opportunities. Not all interventions of equal epidemiological effect have equal evolutionary effect.
The structural insight is that in any system where the target can adapt during suppression, the timing of releasing pressure is not simply a question of “enough” but of whether the environment has changed enough to prevent the adapted target from exploiting what it learned. Partial suppression followed by premature release can be worse than no suppression, because the suppression period is a training period for the thing being suppressed.
(arXiv:2603.18801)