The prepare steps originally stated that:

In addition to protocol-defined requirements, a [=user agent=] might apply additional validation criteria based on local policy, configuration, or evolving security considerations. For example, a [=user agent=] might reject a request that (a) seeks particular credential attributes, (b) uses or requires cryptographic algorithms the [=user agent=] is configured not to accept (e.g., as part of algorithm agility or a transition to post-quantum schemes), or (c) relies on certificates or trust anchors that are not accepted by the [=user agent=]'s configured trust decisions.

It would be good to expand on these.