ControlPoint Manual — versioned with the code; every worked example below is a REAL corpus job and its numbers are pinned by the test suite.
3 — Control & vector check
What it does
Before constraining anything, the free network is fitted rigidly (scale held at 1.0) to the control you select, and every candidate match is listed with its residual — worst first, with a Where column so you can see which end moved. You tick what to hold; the fit re-runs live on your ticked basis. Sideshots can be held as control too — valid (they carry F1/F2 redundancy) but not recommended, and the software says so.
Worked example — 2HR19 on the 2HR07 nails
The traverse re-occupied nails set by the earlier 2HR07 survey. The vector check found the whole set rotated 64.4″ and shifted +4.4 mm in height relative to the stored control — a genuine datum difference between campaigns, visible BEFORE it could contaminate the adjustment.
Theory in one paragraph
Angles and distances define shape, not position: the free network floats. Fitting it to control with scale fixed separates "the survey disagrees with itself" (bad — fix it first) from "the survey disagrees with the control" (a datum question you decide). Holding control then transfers its frame to every measured point.