The XML Schema Working Group discussed this issue at some length during
a face to face meeting at the end of March.  

Note first that the UPA constraint of XSD 1.0 has in fact been
eased somewhat by the introduction of weakened wildcards.

There was some sentiment (at least one member of the WG) for going
further and eliminating the Unique Particle Attribution constraint 
entirely, as being irrational and unhelpful.  But those favoring 
that measure were in a distinct minority.  Others felt
that eliminating the constraint was too big a change for XML Schema 1.1
but that it might be worth coming back to later.  

Some in the WG argued that the UPA constraint does provide some help 
for certain kinds of tools and tool development. In this connection,
it was suggested that where determinism is helpful, it would suffice 
for spec to require that the input/output mapping (or in other words 
the input / PSVI mapping) of a given complex type be deterministic;
UPA is strictly stronger than such a constraint.  (A non-deterministic
automaton may have a deterministic mapping if each pair of competing
particles will provide the same annotations in the PSVI.)  Unfortunately,
we don't at the moment know enough about the closure properties of
finite-state automata which produce output to be confident about
moving toward a constraint phrased in terms of them.

In the end, the chair determined that the Working Group did not have
sufficient consensus to make this change, so we agreed to close the
issue without further action.  Since the proposal to ease the UPA
constraint had active support, we chose to give the issue a resolution 
of LATER, indicating that we recommend to any Working Group
preparing a future version of XML Schema that they consider this
issue anew. 

Accordingly, I'm marking this issue RESOLVED / LATER.