I looked at imContentSink.jsm/convbrowser.xml and studied the Instantbird audit done by Mozilla. Almost all issues mentioned in the audit got fixed; one is left which does not seem to bring a high-risk with it especially, as Tor Messenger is configured to use the least permissive rendering mode (which is further hardened)
ToDo:
look closer at cleanupNode() and change history
look at DOMParser mainly for making sure that no script etc. execution is happening prior to sanitization
look closely at usage of TXTToHTML converter (used in convbrowser.xml, xmpp.js, xmpp-xml.jsm, ircUtils.jsm and imThemes.jsm)
Oh, my comments there simply were that it would be strictly better if we knew the message content window had scripts disabled fully, rather than (or actually, in addition to) being XSS filtered.
I know there was some contention with the instantbird team over this, as they felt this would negatively impact UI, so my thought was that we could have a security slider here, too. So for example, look at all of the stuff that Coy.im disables: https://coy.im/about/. Rather than killing all of that up-front (which will negatively impact UX), it could be progressively disabled as security level is increased.
Thanks for clarifying (actually, I think you were being pretty clear to begin with, I'm just of the confused type).
A security slider seems apt. In #17480 (closed), Tails is asking for a less restrictive (linkified) content window. (Though maybe Tails, and the rest of the security conscious folks, have move on to CoyIM as it is ...)