I think the discussion is drifting in the wrong direction.
As far as I as an individual user of LaTeX am concerned, things
are fine.  If I need something, I can find it in teTeX, if it's
not there, I'll get it from CTAN.  No problem.

The question is only:

What packages are necessary for author-author and author-publisher
document exchange (in all fields in which LaTeX is used)?

I, as the author of a scientific article, want
to be sure that if I stick to certain conventions (which may be
well be much narrower than the whole of CTAN), any reasonable
publisher will process my manuscript without delay (and screwups
due to retyping, the publisher not knowing about amsmath, etc.).
The publisher will probably also want the same guidelines in order
to reject the screwy stuff for electronic submission, thus giving
the author an "incentive" to submit reasonable files.

This, however, cannot work unless a base of packages is defined
(if the base is too narrow, authors will do screwy things out
of desperation), and UNLESS IT IS WELL DOCUMENTED, which, as
far as the practical problems go, is problem number one
(perhaps on the same level as the frontmatter stuff).