Is it a common experience for architects to encounter issues with active/active registrar designs?  Specifically, when clients are the UAS of a dialog?  Some user agents accept INVITEs from any source once registered while others are not as accepting.

There are ways around this using 302 redirect, but there seem to also be ways per RFC to signal trusted nodes to clients.  RFC 3608 seems to be an excellent way to address this issue by communicating trustworthy nodes to the client upon registration.  However, the Service-Route header seems to give options to the client rather than communicate trust.

No where in the RFC does trust seem to be mentioned in its design, however one could imply it should be deduced by the presence of a Service-Route header.

Nonetheless, even if it should communicate trust, it doesn't seem to do so in all cases so the more logical approach to active/active is signal via a 302 redirect where the registration lives, handling INVITE transactions on the server-side instead of client-side.

I'm curious to hear if there is any advice on from the community overcoming trust issues with user agents in an active/active registrar design.