The constraint language described in SICP (AFAIK began as a paper by Gerry Sussman and Richard Stallman, continued in the PhD thesis of Guy Steele) is indeed an inspiration / precursor to Alexey's dissertation on propagators.
Yes, and with Alexey's work (if I've understood it) the propagators are allowed to hold multiple values(beliefs) based on more than one (possibly conflicting) set of constraints.