Ghosts that Connect

The Faddeev--Popov procedure poses two conceptual puzzles. \emph{Puzzle~(1)}: if gauge-equivalent configurations represent the same physics, the quotient $\F/\G$ should suffice to compute physical amplitudes --- yet the procedure requires anti-commuting auxiliary fields, the ghosts, with no analogue on the quotient. What structure of $\F$ do they encode? \emph{Puzzle~(2)}: gauge-fixing was supposed to eliminate local gauge symmetry, yet the gauge-fixed theory retains BRST --- a residual symmetry that acts on the gauge potential as an infinitesimal gauge transformation. Why does it survive? Both puzzles dissolve together. Following \textcite{Dougherty2021} and \textcite{DoughertyRead2026}, I take ghosts to encode classical content of $\F \to \F/\G$, but identify a different structure: a principal connection $\varpi$ on this bundle. The ghost is $\varpi$; the BRST operator is the vertical exterior derivative on field space; the Maurer--Cartan equation is its vertical Cartan structure equation. The Faddeev--Popov calculus draws only on $\varpi$'s vertical content, which the algebraic reading also captures; $\varpi$'s horizontal content, on which the Vilkovisky--DeWitt programme rests, supplies the cross-orbit pairing that gauge-fixing, dressing-based quantisation, and counterfactual comparison require and the quotient discards. BRST is the rigid, vertical symmetry that preserves this pairing; this is why it survives gauge-fixing.