diff --git a/PLAN.md b/PLAN.md
index 1dc4741..b23a90c 100644
--- a/PLAN.md
+++ b/PLAN.md
@@ -2249,17 +2249,18 @@ NPCs* step to the profiles page. The canonical documents gain `runicnpc/API.md`,
 `INSTALL.md` and `PLAN.md` under Rust, and that page now names the API page as the one deliberate
 exception to "every page links out".
 
-**Findings outside this repository**, reported rather than fixed:
+**Two findings outside this repository**, reported in the PR:
 
-- **The Rust bundle's Oxide floor is lower than RunicNPC's.** `v2/rust/current.json` says Oxide
-  2.0.7585, which is the bridge plugin's floor. But the bundle requires RunicNPC, whose own
-  `plugin.toml` says 2.0.7726 — and its comment notes that RunicNPC "has never run on" 7585. So
-  the installer will accept a server that RunicNPC has never been tried on. The site states both
-  numbers where each belongs (the Rust overview quotes the bundle's, the RunicNPC pages
-  RunicNPC's) and claims nothing about the gap.
-- **`API.md`'s profile table breaks in two places.** The `boss` rows and the `loot` rows each
-  follow a blank line with no header row, so Gitea renders them as plain text full of pipes rather
-  than as tables. The check parses them line by line and is not affected; a reader of `API.md` is.
+- **The Rust bundle's Oxide floor is lower than RunicNPC's — and that is intended.**
+  `v2/rust/current.json` says Oxide 2.0.7585, the bridge plugin's floor, while RunicNPC's own
+  `plugin.toml` says 2.0.7726. The org lead ruled on 2026-10-22 that **2.0.7585 is the original
+  compatible version and stays**. It moves only when the CI staging drill reports an issue, and then
+  as part of the update that fixes it. The site states each number where it belongs (the Rust
+  overview quotes the bundle's, the RunicNPC pages RunicNPC's) and claims nothing about the gap.
+- **`API.md`'s profile table broke in two places.** The `boss` rows and the `loot` rows each
+  followed a blank line with no header row, so Gitea rendered them as plain text full of pipes.
+  The check parses them line by line and was not affected. Fixed at the org lead's ask in `docs`
+  #371, which removes the two blank lines and changes nothing else.
 
 ---