docs(diff): describe why the post-render chain is safe, accurately

The previous comment claimed this effect is declared before every consumer's effects so each emit
sees current values. Both halves are false: it is declared after them, and a child would win
regardless -- Pierre renders from its own layout effect and React commits children first, so an
emit can read the previous cycle's callbacks.

What actually makes it safe is that FileDiff.render early-returns without emitting when the diff,
annotations, theme and range are unchanged, so search-only and note-target-only updates never
reach a stale chain. Record the one real overlap -- content changing while a search is active --
and that it self-corrects on the next results tick.

src/renderer/src/components/editor/pierre-diff/PierreDiffSurface.tsx

@@ -176,8 +176,15 @@ export function PierreDiffSurface({
   })
   // Why layout and not render: Pierre treats a new onPostRender identity as forceRender, so the
   // callback below must stay stable and read its chain from here. Writing during render is impure
-  // -- React can discard that work. useRef seeds the first render's chain, and this effect is
-  // declared before every consumer's own effects, so each Pierre post-render sees current values.
+  // -- React can discard that work.
+  //
+  // This does NOT guarantee a fresh chain on every emit, and it cannot: Pierre renders from its
+  // own layout effect in a child, and children commit before parents, so an emit can read the
+  // previous cycle's callbacks. useRef seeds the first mount. What keeps that harmless is that
+  // FileDiff.render early-returns without emitting when the diff, annotations, theme and range are
+  // all unchanged, so search-only and note-target-only updates never reach a stale chain. The one
+  // overlap -- content changing while a search is active -- can paint a frame of stale ranges; the
+  // next results tick corrects it.
   useLayoutEffect(() => {
     postRenderRef.current = {
       onPostRender,