I'm not sure it would be the same problem if you are running it on a failover cluster instance. On a traditional cluster (i.e. AlwaysOn FCI) if you need MSDTC that becomes a clustered resource. So if the cluster fails over to another node, the MSDTC resource would fail over as well, so I think any transactions that were not yet committed would roll back. I would think the issue with MSDTC and an AlwaysOn Availability Group is that the AG is cluster-aware, but MSDTC is not. So if the AG fails over there is nothing to rollback the other parts of the transaction on other databases.

At least that is what I would think the issue would be.