I agree with my friend Bob Stern that the onset of actualistic plate tectonics is an unsolved problem, and that the answer must ultimately come from geological evidence and not from model calculations. There are simply too many unknown factors (e.g., eclogitization, oceanic crustal delamination) that could have affected lithospheric buoyancy in the early Earth.
There are, however, diagnostic products of subduction not considered by Stern (2005). Arc magmatism is genetically linked to lithospheric subduction and is distinctive in type and distribution. Arc magmatism occurred throughout the Proterozoic and Archean (Hoffman, 1989). The example I am most familiar with is the Great Bear magmatic zone (1875–1845 Ma) in the northwest of the Canadian Shield. This linear 1000-km-long by 100-km-wide zone was originally interpreted as a magmatic arc related to east-dipping subduction on the basis of petrology, petrochemistry, geochronology, aeromagnetics, and tectonic setting (Hildebrand et al., 1987; Hoffman, 1987). Deep seismic reflection profiling later revealed east-dipping mantle reflections at a depth of ~100 km beneath the zone, which “almost certainly represent a subduction surface associated with arc development” (Cook et al., 1999, p. 1). They are among the strongest mantle reflections of any age. Equally significant is the tectonic setting of the magmatic arc, which initiated ≤7 m.y. after the collision of an exotic island arc with a west-facing passive continental margin (Hoffman, 1980;
Hildebrand et al., 1987; Bowring and Grotzinger, 1992; Gandhi and van Breemen, 2005). It shows that the Andean-type magmatic arc developed as a consequence of arc-continent collision and subduction polarity reversal. This process, presently underway in Taiwan (Suppe, 1984), has long been recognized as the fundamental means through which passive margins become active margins (McKenzie, 1969, their Fig. 13). The Great Bear magmatic zone is just one of countless former magmatic arcs indicating lithospheric subduction long before the Neoproterozoic. Moreover, Archean boninites (Kerrich et al., 1998), boninitic komatiites (Parman et al., 2003), and ultrahigh-pressure crustal xenoliths (Schulze et al., 2003) indicate that subduction began much earlier.