Record the random tree's padding nodes and span add/remove scenarios

docs/IRIS_EXTRACTION_HANDOFF.md

@@ -9,7 +9,7 @@ Canonical `main` is **`ca2b4b2`** (#17, the headless rig). Sixteen slices are
 in; #16's size work and #17's rig both merged on 2026-09-14.
 
 **#18 `split/18-position-chain`** is open, worktree `/home/bob/repos/iris-pr18`,
-head `2272634`, ten commits. LAYOUT.md §2's O(1) subtree movement, the
+head `4178dfb`, eleven commits. LAYOUT.md §2's O(1) subtree movement, the
 `Remap` retirement, and then the three changes the owner chose on 2026-09-14:
 the slot carries a box, slots are opt-in, and a widget's region is held in the
 coordinates of the slot it draws in.
@@ -91,12 +91,24 @@ A live resize does land where a cold start does, byte for byte, on `tabs` and
 compositor and a single output, so two at once resize each other's window and
 screenshot the wrong thing.
 
-**Rigs added.** `iris::random` grows the tree (spans in every direction,
+**Rigs added.** `iris::random` grows a seeded tree -- spans in every direction
-stacks, rects with varying opacity, text wrapping and overflowing, a declared
+holding two to four children, stacks, padding with each of its four sides its
-size over half of it, stopping at a depth), `examples/random.rs` draws one --
+own number, rects with varying opacity, text both wrapping and overflowing, a
-`IRIS_SEED` and `IRIS_DEPTH` pick it -- and `tests/generated.rs` compares warm
+declared size over half of it, stopping at a depth. `examples/random.rs` draws
-against cold under a size change, a resize, and both. It found the
+one (`IRIS_SEED`, `IRIS_DEPTH`), and `tests/generated.rs` grows each seed twice
-non-terminating marking above immediately.
+-- once and then changed, once with the change built in -- and compares every
+widget's box. Eight scenarios: a size change, a resize, both, and five ways of
+changing what a span holds (every other child out, all but the first out,
+three on at once, the first out and three on, one out of the middle and one on
+the end). The ignored sweep is 100 seeds across all eight, 800 comparisons.
+
+Two things it needs that are easy to get wrong. Three spare leaves are grown
+beside every span whether they end up in it or not, and detached children are
+held until the comparison is over: a tree that makes fewer widgets, or frees
+one whose id is then handed to the next, stops lining up index for index and
+every comparison after the first difference is against the wrong widget. And
+each shuffle asserts the tree actually changed before comparing, so a case
+that quietly did nothing fails rather than passes.
 
 ### The plan after #18, agreed with the owner on 2026-09-14