back to statistics and biostatistics
even if we have a randomized trial, it might be better to adjust for covariates
https://scholar.google.ca/scholar?q=efficiency+covariate+adjustment+randomized+trial&btnG=&hl=en&as_sdt=0%2C5&as_vis=1
I didn't read more than abstracts.
In the inverse probability weighting literature there was an earlier result that it's more efficient to use an estimated propensity score than a known propensity score (like randomization).
Looks like a "hot topic" where the literature for treatment effects for observational studies comes back to randomized trials.
(Just guessing)
Josef