I would propose that we go for the option of letting OpaqueBehaviors being redefinable. Since we have
bug 511060 in the pipe for 1.0, related to the handling of the RTGuard in the model import tool, we can have one more bug done at the same time for the handling need to establish the redefinition assocation for, including applying the RTRedefinedElement stereotypes on, OpaqueBehaviors during import. Should hopefully not be that much more work if both are done at the same time.
One benefit for end-users I could foresee with doing this in the "right way", is that the proposed improvements to EMF Compare that Philip mentioned the other day, see
bug 512514, regarding proper support for redefinition, hopefully can improve the way truly redefined OpaqueBehaviors can be presented in EMF Compare. Apart from that I doubt that any end-user really bothers about a correct redefinition being established or not. What do you say Philip, do you think that the proposed improvements to EMF Compare could be useful also for this case as well?