Classify infrastructure short-span outlier families

docs/control-loop-atlas/runtime-roots-camera-and-support-families.md

@@ -2980,6 +2980,18 @@ The low helper strip beneath that shared family is tighter now too: `0x0052ecd0`
   while the zero-length class is a separate `0x0055 / 0x00` outlier across `18/18` rows and the
   `0x06` class is the only large mixed frontier left. So the next infrastructure pass should focus
   on classifying the mixed `0x06` rows instead of re-proving the pure-prelude `0x03` class.
+  That `0x06` class is now narrower too: grounded `q.gms` shows the dominant short-span family as
+  `BridgeSTWood_Section.3dp / Infrastructure` with compact prefix `0xff000000 / 0x0001 / 0xff`
+  across `62/72` rows and dominant prelude candidate `0x0001 / 0xff` across `63/72` rows. So the
+  next infrastructure pass should stop treating all short rows as equally ambiguous and focus on
+  the smaller outlier families inside that class, especially the `BallastCap`-style zero-like
+  rows and any remaining non-`0x0001 / 0xff` prelude candidates.
+  Those outliers are explicit now too: the remaining `10` short-span rows on grounded `q.gms`
+  break into `9` `BallastCapST_Cap.3dp / Infrastructure` rows with compact prefix
+  `0xf3010100 / 0x0055 / 0x00` and candidate pattern `0x0055 / 0x00`, plus `1`
+  `TrackCapST_Cap.3dp / Infrastructure` row with compact prefix `0xff0000ff / 0x0001 / 0xff`.
+  So the next infrastructure pass should target the `BallastCap` outlier family first instead of
+  spending time on the already-dominant bridge-section class.
   The child loader family is explicit now too: local `.rdata` at `0x005cfd00` proves the
   `Infrastructure` child vtable uses the shared tagged callback strip directly, with
   `+0x40 = 0x00455fc0`, `+0x48 = 0x00455870`, and `+0x4c = 0x00455930`. So the remaining