Only requirement is that the domain the attacker is spoofing is using O365.
This is not true. The paper mentions multiple service providers using more relaxed validation.
Table 3, section 5 in the paper shows which policies need to be in place on the domain they are piggy-backing on.
They reference Postfix:
"Additionally, we note that mailing list software such as Listserv and Mailman require a backend MTA. In our experiments we used Postfix with DMARC turned on, a configuration which follows good security practice. However, in practice many organizations might not use this configuration because many MTAs (including Postfix) do not enforce DMARC by default. In these cases, the attacker can spoof email from any target domain, regard- less of its DMARC policy, much like the attack against Gaggle."
I read this to mean that if you actually enable DMARC in Postfix, piggy-backing on another domain's policies results in rejection.
No mention of results for receiving at ProofPoint, Mimecast, Trellix, or Cisco's email appliance.
This is not a UX problem.
They are demonstrating a problem with managed providers, and their opinionated configuration. You give up a lot of control as an admin when you use 365 as your front-end. This further proves that.