diff --git a/operator-token-runbook.md b/operator-token-runbook.md
index a66c4d4..0d9b9b6 100644
--- a/operator-token-runbook.md
+++ b/operator-token-runbook.md
@@ -65,3 +65,33 @@ Env vars for custom apps are set via:
 
 - Cloudron CLI: `npm install -g cloudron` (install on your workstation, not the server), then `cloudron login my.inference.coop`.
 - Cloudron API docs: `docs.cloudron.io/api/`.
+
+## Open Collective email visibility can regress (Sept 2026)
+
+The portal identifies members by email, read via the bot's OC personal token
+(`OC_PERSONAL_TOKEN`) as `collective.members.nodes[].account.email` with the
+`... on Individual { email }` inline fragment. This is normally populated
+because the bot is an **admin** of the collective.
+
+**Failure mode:** the GraphQL `email` field can flip to `null` for *every*
+member — including the bot's own `me { email }` — while:
+
+- the bot still authenticates and reports `isAdmin: true`,
+- the OC **admin CSV export still shows emails**.
+
+When this happens, `fetch_members()` filters out empty emails and returns `[]`,
+so the webhook can't provision anyone by email and the nightly sweep goes into
+its empty-list fail-safe (skips, doesn't deactivate). Symptoms: a new member's
+webhook fires cleanly but no "Sent email to …" log line follows, and no
+Cloudron user is created.
+
+**Diagnosis:** run `me { email }` with the bot token. `null` here (while
+`isAdmin: true`) confirms the regression is OC-side in the GraphQL layer, *not*
+a token rotation and *not* a lost admin grant. The CSV is the independent
+ground-truth check.
+
+**Interim workaround:** for any member who can't be provisioned by email, get
+their address another way and onboard via the manual path — add to
+`MANUAL_MEMBERS`, then `POST /admin/provision` (X-Admin-Token header). The
+guest-with-no-email case is *not* the cause; verify with `me { email }` before
+concluding a member is unreachable.