Progress is staged, not redefined

agent-decided, reversible. The question of whether Progress should be claim-weighted or service-weighted had a false disjunct. The base theorem stays service-weighted and unchanged; a second theorem, whose premises are a transport plan and its constants and which mentions no engine, carries the conclusion to the claim-weighted one at a residual equal to the transport stability error.

The argument is premise locality. Stage one mentions no scheduler and no claim measure; stage two mentions no gain, no margin, no Uptake and no engine, so it composes with any engine-side guarantee of the form E_nu[d] -> 0 — including the merged Progress schematic exactly as settled, and including the Logical Induction realization, which reaches that form by a route sharing no premise with it. The round's two central countermodels sort the same way: the rotation countermodel is entirely a stage-two failure and the vanishing-share countermodel entirely a stage-one failure. In the merged form the residual sits in a theorem whose other premises are semantic and learner-side, so a reader cannot tell what it is charged to.

No settled definition changes, and the question leaves Awaiting the author because it has an answer rather than because it was ruled on.

Rejected alternative: rewriting the base Progress theorem to be claim-weighted, which would import a scheduling object and a residual into a statement that needs neither.