For a decade, the plasmoid instability was the leading explanation for fast magnetic reconnection — the explosive release of energy that powers solar flares and tokamak disruptions. The theory: when a current sheet thins sufficiently, it fragments into chains of magnetic islands (plasmoids) that coalesce in a cascade, accelerating reconnection to rates independent of resistivity. The mechanism was elegant, predictive, and confirmed by every simulation that tested it.
Those simulations were not resolved finely enough.
At grid resolutions up to 65,536², the plasmoid picture disintegrates. For Lundquist numbers up to S ~ 10⁴, reconnection follows slow Sweet-Parker scaling. Between 2×10⁴ and 2×10⁵, plasmoids do form — but they are rapidly advected out of the reconnection layer before they can merge. The cascade that supposedly drives fast reconnection never completes. The result: V_rec ~ S^{-1/3}, still resistivity-dependent, still slow.
Only above S > 2×10⁵ does reconnection saturate near V_rec/V_A ~ 0.01. But at those parameters, Reynolds numbers exceed 2000, meaning turbulence and three-dimensional effects dominate. The 2D plasmoid mechanism, operating alone, cannot explain fast reconnection in astrophysical settings.
The plasmoids were always real. What was wrong was the attribution. Earlier simulations lacked the resolution to see the plasmoids being swept away — at lower resolution, the advection was artificially suppressed, allowing the cascade to proceed. The mechanism existed in the simulations but not in the physics.
When a theory that works at low resolution fails at high resolution, the theory was describing the simulation, not the phenomenon. Some mechanisms survive only in approximation.