We are using relative timeouts with interceptors. On the logs timeouts seem to be reached but on reality they don't.

Analysis of the problem:
We are using RELATIVE_REQ_TIMEOUT_POLICY_TYPE policy by code (Orb.create_policy() method).
Within interceptors we call another corba services. For instance we call the jacorb naming service, calling the resolve method on _NamingContextStub (JacORB IDL compiler V 3.2).
This has the effect to call the Delegate.request() method with key "interceptor_call" on the thread context and to set clearCurrentContext attribute to false. A main error is that, clearCurrentContext attribute is never set to true later (and clearCurrentContext should be an attribute of the current thread), and as Delegate is a singleton class, some independant calls of request method by other class in another threads should not clear the thread context on invoke_internal methods (and time end resquet infomations). Then on new call of request method, the request end time still exists on the thread context and is not replaced by a new one computed. So with ellapsed time, because these request end times are never removed or replaced, finally we get allways exceptions on line 2209.