MF plugin (specifically most of the Octane CI integration logic) has several Jenkins' events listeners - on builds, on workflow, on SCM and on item changes.

Those listeners are running on the mail thread, and therefore any uncaught failure in them will cause the whole build to fail.

This is not correct. Correct way to do it, is that any non-mission critical failure (which is none of the Octane's plugin provided functionalities as of now) should handle any failure silently, log it to its own log and let the build to continue.