Subduction megathrusts generate great earthquakes and tsunamis. Understanding what controls their seismogenic zone has important science and hazard implications. In the 1990’s, Canadian Geophysicist Roy Hyndman championed a thermal-control hypothesis which guided global research for many years. In this hypothesis, clay minerals stable up to ~150C defines the updip limit of the seismogenic zone, and the serpentinized mantle wedge corner (MWC) or a temperature of ~350C, whichever is shallower, defines the downdip limit. Today, modern observations show that many megathrust ruptures extend much beyond these limits. However, here I explain that the core concepts in the Hyndman hypothesis are still of great value and provide basis for a new conceptual framework that incorporates findings of rupture dynamics and fault slip behaviour made in the 21st century. These findings indicate that a fault segment that impedes seismic slip can participate in the slip until gaining sufficient strength to stop the slip, or it may exhibit dynamic weakening to facilitate the slip if a high-enough slip rate is attained. The clayey shallow segment and serpentinite-rich MWC segment of the megathrust thus act as such “soft barriers” to seismic slip. The 21st-century findings also indicate that high-temperature, antigorite serpentinites exhibit seismic behaviour, explaining why seismic slip is often seen to propagate to, or even initiate at, depths 10-20 km deeper than the MWC. For very warm subduction zones such as Cascadia, the discovery of Episodic Tremor and Slip (ETS) at the turn of the century adds a new dimension. In these subduction zones, the megathrust seismogenic zone is limited to a shallower depth than the MWC by thermally activated creep. The thermal and petrologic conditions at the MWC harbour ETS instead of a soft barrier to seismic rupture.