I don't think there is much hope for getting such folklore
documented, since there is a serious embarassment issue to
consider. Much of this folklore involves dealing with bugs
that should have become extinct years ago. A significant
subset of these consist of problems that could have been
readily addressed by minor modifications to the protocol
specification, but which for political reasons never
happened.
For example, RFC-822 implementations which expect the BNF
for `specials' to work as documented are in for a shock in
the real world. You end up with two sets of `specials'; one
which you use to generate RFC-822 and another which you use
to parse. The same thing is happening now with MIME, in
spite of my (evidentally futile) attempts to nip it in the
bud.
The old saw of `be liberal in what you accept, be
conservative in what you generate' hides a great many sins.
It is easier, I think, to recite this than it is to admit to
our past (and present) mistakes.