I'm using rspamd integration from early versions of beta releases. It works great. I even moved my custom spamassassin rules (they are very basic) as a plugin to rspamd and it works.

Messages are of course not being greylisted, rejected or nothing is being added to header by rspamd but score is being returned to SmarterMail which is being used for spam classification. rspamd is using A LOT less memory and cpu than spam assassin. When I upgrade my production server with client data to latest SmarterMail then I'm reeeeaaaaalllyyyy interested how marking messages as HAM/SPAM by users from webmail will work with teaching rspamd filters what is spam and what is not.

EDIT: From early versions -> I use it as incoming gateways.