I re-read the Phoenix Project recently and it was interesting to reflect on things. The Phoenix project was late, slow, and not working. So they started the Unicorn project to address the most critical things that weren't getting done, essentially re-platforming the Phoenix project while it was still not finished.
It was more of an evolution of the project to the side, which I think makes the most sense. Even with something like a mainframe, adding to the side of that with newer architectures/tech while some maintenance/dev happens in that space, is often a better idea. I wouldn't lift a current on -prem app to the cloud, I'd start moving pieces of it as a new app(s) to evolve the architecture rather than replace.