Turns out he probably wasn't making Roman concrete (just a bad modern imitation) as recent discovery suggests a vital component of their mix was hot mixing/using quicklime - https://news.mit.edu/2023/roman-concrete-durability-lime-casts-0106 - which results in self-healing/protective lumps which react to water ingress to protect the whole structure. In my understanding this is kind of like what he did in the initial (prep of ingredients) steps but actually during the mixing (which he did cold). Interestingly this results in a much less consistently mixed result -which had been written off as bad mixing.
Abusive comment hidden.
(Show it anyway.)