Plastic faulting is a brittle‐like failure phenomenon exhibited by water ice and several other rock types under confinement. It is suspected to be the mechanism of deep earthquakes and extreme cases of shear localization in shallow rocks. Unlike ordinary Coulombic failure, plastic faulting is characterized by a pressure‐independent failure strength and fault plane oriented 45° to maximum principal stress. To research the question of how the instability initiates, we conducted over 50 constant‐displacement‐rate experiments on polycrystalline ice (phases Ih and II) near the brittle‐to‐ductile (B‐D) transition, at confining pressures P = 0–300 MPa, applied strain rates  = 5 × 10⁻⁵ – 7 × 10⁻³ s⁻¹, temperatures T = 105–233 K, and mean grain sizes d = 0.25–1.18 mm. We find that (1) the width of the B‐D transition in variable space is vanishingly narrow, to the point of appearing as a crossover, (2) a plastic fault plane, once formed, is not a zone of subsequent weakness, (3) distributed ice I→II phase transformation in small amounts (<1 vol%) shows no causal relationship to subsequent failure, and (4) plastic faulting also occurs in ice II. We hypothesize that the elusive nucleating “trigger” parallels that of metals and ceramics undergoing severe plastic deformation, wherein transient local structural rearrangement occurs, in turn causing material strength to drop to a level sufficiently low, in a volume sufficiently large, that adiabatic instability is nucleated. Our results do not require and often are inconsistent with phase transformation. Plastic faulting may therefore be available to all solids undergoing severe deformation, and its appearance in so few is simply the result of insufficiently extreme conditions.