Still, I don't see the ALLOCATABLE components problem as all
that bad.  The result of TRANSFER must be assigned or it's lost.
(OK, it could also be an actual argument to a procedure, and maybe
there are other uses I can't think of on an Easter night.  But let's
consider
the obvious first.)  So, suppose you do:

   var = transfer(transfer(E, D), E)

Where E and VAR are both some derived type and D is some type
big enough to hold all the bits of the derived type.  In this case, the
ALLOCATABLE components of VAR do *not* end up associated
with the same storage as the components of E.  This is because of
the semantics of assignment.  It has nothing to do with TRANSFER
at all.  It's the same result as if you had written var = E.  It may be
that the other cases, such as passing the result of TRANSFER as an
actual argument to a procedure, turn out to be equally benign.  You
would have to go through each possibility carefully.  If they aren't
all benign, well that's one of the kinds of havoc.

--
J. Giles