Deligne's proof that inverse image functors preserve flat modules is to internalize the Govorov-Lazard proof that a module is flat iff it is a projective colimit of finite free modules, and to show inverse image functors preserve these internalized colimits (SGA 4, Springer-Verlag edition, Exp. V appendix). The classical proof (say, Eisenbud COMMUTATIVE ALGEBRA appendix 6, theorem A6.6) looks good for topos logic though I admit I have not worked through internalizing it. Also, I have still not fought through the definitions in Deligne's appendix to SGA 4. I would still appreciate any reference to a cleaner treatment if anyone knows one. It seems likely that Deligne's "local inductive limits" are internal colimits in the modern sense but I have not read the thing closely enough yet to really say that. This seems to me a strikingly sophisticated lifting of a theorem, from SET to any topos, for 1972. The current methods for routine lifting were just being created over here in elementary topos theory at that time. I believe Deligne did not know of that. Or am I making too much of this proof? I'd appreciate any historical reflections on relations between such lifting in the Grothendieck school, and methods of elementary topos theory. Colin