Need to clarify from the OMG/one of the CORBA spec experts the semantics/use of
the CORBA::Policy::destroy method.  In particular, whether the user/app.
developer is allowed to call <destroy> method on policies returned to him from
CORBA::PolicyManger::get_overrides method.  This is important for the correct
implementation of TAO_Policy_Manager_Impl, which keeps track of all set
policies.
In the current implementation of TAO_Policy_Manager_Impl, if destroy method is
invoked on one of the policies it stores in the Policy list, any consequent
operation on TAO_Policy_Manager_Impl, involving the deleted policies would
result in a seg-fault (because policies have direct collocation scheme).
Again, we need to find out what the correct semantics is in terms of users
invoking <destroy> on policies in the PolicyManager, and try, if possible, to
throw exception/provide error message if the user performs any illegal
operation, rather than segfaulting at a later point.