They're very relevent questions.
I think that you must either do a painstaking package by package
upgrade, or take advantage of someone else's work at doing the same.
Luckily the chicken-egg problems are starting to disappear as
everything is gearing up for a stable release cycle, or is it really
getting worse and I somehow skirted through it all?
I see the purpose of the stable Linux version as a rallying point
around which all the other package maintainers stabilize so that
complete systems integrators may put together decent distributions,
which you in turn get and install (following the procedures as
outlined by the distribution you pick).
However, as to the technical question of whether Linux 2.0 will work
with an environment made around the date of the 1.2 stable kernels, I
do not know but suspect that the answer is somewhat difficult to
really say (e.g., proc-ps must be upgraded; does it require a later
libc? In so doing, does that require new package reconstruction at
every level?) Perhaps someone ought to try it out and post their
results, but I wonder, is there a use to this approach? If 1.2 works
fine, why not just use it, and if not, then why not just do a full
upgrade?
I don't know answers but want to support you in your question.
I for one find Documentation/Changes extremely useful in keeping
current, but it doesn't exactly deal with your question directly.