docs(drift): a mechanism inferred on the wrong axis, parked

[JulienMartel]

What

One entry appended to docs/drift.md's Seen once, not yet a row. Row numbering untouched — still thirty-three.

Why

The largePrint lane's handoff carried a mechanism from ops#69: what haus renders as a file comes back, what lands in macOS's own preference store survives. It fits every observation it was drawn from, and it names the wrong axis.

NSTableViewDefaultSizeMode is a plain NSGlobalDomain preference that the same haus.ui.scale moves, and it returns — because haus writes it at every scale, where dock.tilesize in the same block is written under mkIf (ui.scale != 1.0). What predicts is whether haus writes the setting at both values of the switch; a rendered file is simply the case that is always rewritten.

The unit shifts with the axis, which is part of what hides it: counted per option the split is two back and two not, and ui.scale sits on both sides of it the moment you count writes instead.

Parked rather than numbered: seen once.

Verify

The entry disposes of rows 5, 18 and 26 by name — the half of the section's stated bar ("or tell it from a row already here") that is easy to skip. The counterexample it rests on is now evaluated in haus by largeprint-return (hausfold/haus#760), and was already visible, unnoticed, in scale-reach's adjacent sys dock.tilesize / sys finder.sidebar rows.

Watch out

• It cites identifiers and files, not line numbers — haus/modules/core/default.nix moves under it in appearance: largePrint says which of its four leaves false puts back haus#760.
• Its Caught by points at a table that already existed. That is the finding, not an oversight: the cheap move is to read the seam you already keep, at the cell where two candidate axes disagree, rather than build a new one to go looking.