i said:
>   thousands/hundreds/dozens

and just to remind everybody, the solution is very clear,
and has been for a very long time, ever since we learned
-- by my careful analysis of the many d.p. experiments --
that p1 proofers find as many errors in subsequent proofings
as p2 proofers and even p3 proofers do.  (indeed, of the 3,
the p2 proofers were the least good at locating the errors.)

so it's obvious that we can move text to perfection by simply
running it through p1 repeatedly.  one problem with that --
as we've already found -- is that sometimes p1 proofers will
change correct text to incorrect text.  that problem can be
eliminated easily with a policy to review and reconcile diffs.
(this policy is easy to implement roundlessly, and will also
serve to train up your low-quality proofers, so it's win-win.)
the other problem currently with repeated p1 is that d.p.
hasn't created an unambiguous set of proofing instructions
-- i know, you'd think the need for that would be obvious --
and thus sometimes proofers "cycle through" corrections...
(e.g., a first proofer dehyphenates, a second rehyphenates,
a third asterisks the hyphen, a fourth dehyphenates, etc.)

i haven't discussed the f1/f2 problem, because it's a mirror
of the p1/p2/p3 problem.  a quick-and-easy confirmation
of the f1 by a subsequent f1 view, and we're off to the races.

likewise, i have not discussed the postprocessing problem,
not explicitly for the most part, because it's the microcosm
of the thousands/hundreds/dozens problem, it certainly is.

and once again, the problem is the workflow.  by the time
all of its pages have been proofed and formatted, the book
should fall in place more or less naturally and automatically.

the fact that it does not, in the d.p. workflow, is a shortfall...
it indicates that the workflow is deficient in some major way.
but correcting the postprocessing problems is relatively easy.
you simply need to analyze each page that needs "finishing"
and determine how the proofers could've provided that for it,
and you modify the proofing instructions appropriately. easy.

-bowerbird