Syntactic Ergativity?

Okay, I'm still working on the development of some secondary langs related
to Yivríndil.  Thanks to everyone who gave comments earlier about
palatalization, and now I've got a new problem involving relative clauses
and ergativity.

In the ancestor of Praçí, there were surface constraints about the word
order in relative clauses to avoid ambiguity and for prosodic
concerns.  The rules were:
  Unmarked word order is SOV
  The relative pronoun remains in situ (doesn't front)
  The order SO must be preserved
  Avoid possible ambiguous parsings at all cost
  Sentence elements are arranged to avoid consecutive phrases of the same
type, i.e. no consecutive accusative phrases, etc.  The relative pronoun
itself isn't affected by this, though, so that an accusative pronoun can
occur next to an accusative noun phrase.

All of the possible variations are listed below.  The relative pronoun is
marked RS or RO, depending on whether or not the relative pronoun in an
object or subject in its clause; S, V, and O have their usual
meanings.  The relative clauses are in
brackets, so all of the possible variations come out like this:

Transitive main verb and subordinate verb:
S [RS O V] O V
S [V S RO] O V
S O [RS V O] V
S O [S V RO] V

Intransitive subordinate verb, transitive main verb:
S [RS V] O V
S [V RS] O V
S O [V RS] V

Transitive subordinate verb, intransitive main verb:
S [RS V O] V

Intransitive subordinate verb and main verb:
S [V RS] V

Now, these rules created a lot of strange word-order conventions, so they
were quickly abandoned.  If you look, when the subordinate verb is
transitive the word order RS V O dominates, but when the subordinate verb
is intransitive the word order V RS dominates.  Those two became
generalized for all relative clauses, and then the case markings were
lost, so only word order remained to distinguish subjects from
objects.  But the subject of an intransitive verb *follows* the verb,
which is the same word-order position as that of an object for a
transitive verb: Ergativity, only in relative clauses, and only marked by
word order!

I may generalized this further and allow the word order constraints for
relative clauses to be generalized for all clauses, especially since
non-inflectional languages tend towards SVO anyway.  This would
bring semantic ergativity into the whole language.

Anyway, is this too farfetched?  Has any real language ever developed
ergativity out of a relative clause?  Are there any real languages that
mark ergativity only through word order?  Looking for info

Jesse S. Bangs jaspax@u.washington.edu
"It is of the new things that men tire--of fashions and proposals and
improvements and change.  It is the old things that startle and
intoxicate.  It is the old things that are young."
-G.K. Chesterton _The Napoleon of Notting Hill_