Wallace was still revising between the advance galleys and the first printing…

This concerns the book's apparatus (its title, endnote system, or text-state) rather than a single scene.

In the galley proofs. The 1995 galley proofs = the bound ARC, already typeset at the published pagination ("The galleys are still 1,079 pages long"), sent to reviewers with "a piece of Little, Brown stationary on which was printed one erratum" — i.e. the "712,000"-typos state Wallace joked about was pre-galley, and the ARC was already substantially corrected.

In the published book. This concerns the book as a whole rather than a single scene.

Evidence. A single source reports this. (the only published collation of this layer). Documented by the DFW Words galley-versus-first-printing collation — "Wallace had clearly worked on the book between the advance galleys and the hardcover release"; "it's tempting to look at the galleys as a kind of penultimate draft"; "These are not substantial changes" (word/punctuation grain, no new paragraphs); scope limit stated in-source: "There were additional changes between the hardcover and paperback; I'm just comparing the galleys to the hardcover.".

Bears on. This difference bears on the following open questions: the post-1993 stratum (YoG opening, mold scene, Marathe/Steeply, AFR superstructure) — new canonical slug for claims previously filed ad hoc.

Further notes. Extends the late-additions record one layer past the 1994–95 cut-negotiation: authorial revision continued into the ARC→print window (late 1995), after typesetting. The ~205 decoded variants live in the blocks below; hardcover→paperback variants are OUT of this source's scope by its own statement.