According to the spec (https://confluence-public.open-xchange.com/display/CoiW/COI+...) clients are expected to ask the IMAP server using a CAPABILITY command whether it supports COI extensions. Whether or not the server does support COI, messages are indeed standard mails with an extra Chat-Version: header and the requirement that Message-Ids start with "coi$".
So an existing server need not do anything to support COI clients. But a malicious server could easily decide to reject them, or to filter out COI-specific stuff from messages. I don't think it's likely, but it sometimes big corporations and governments get paranoid about weird things ("network load!!1!") and take technical measures to block them.
I wouldn't be surprised if COI encounters issues with automated abuse detection and spam filters in those servers. When you're in an active conversation, the client is rapidly sending 1-line emails with strange headers.
So an existing server need not do anything to support COI clients. But a malicious server could easily decide to reject them, or to filter out COI-specific stuff from messages. I don't think it's likely, but it sometimes big corporations and governments get paranoid about weird things ("network load!!1!") and take technical measures to block them.