Accepted answer
What the data supports is narrower than what gets recommended, so it is worth separating the two.
DEXA precision is better than people assume for fat mass and worse than people assume for lean mass in a single scan — the least significant change for regional lean mass on a well-maintained scanner is on the order of a few per cent. That means two scans three months apart can differ without anything having happened, and it means a scan sequence needs to be at least three points before a trend is interpretable.
Worth being precise here: the regain trajectory after stopping is roughly a mirror of the loss trajectory, and it is not primarily a willpower phenomenon. Appetite signalling returns, energy expenditure remains suppressed relative to the original mass, and the two combine. That is an argument for a maintenance plan existing before the stop, rather than an argument against stopping.
SURMOUNT-4 provides the cleanest maintenance-versus-withdrawal contrast available in the class, and it is the reference for any claim about what happens after stopping[1].
One qualification: none of this is a clinical assessment, and unexplained loss of function rather than of mass is a reason to see someone rather than to adjust a programme.
Two resistance sessions a week and a protein target you actually hit will do more than any refinement beyond them.