docs: mark MX-exclusion plan complete (all 3 fixes shipped)

Fix 2 (untagged NULL bucket cap) shipped in bc93d93; default is no-starve.
Plan fully implemented.

docs/plan.mx-exclusion-gaps.md

@@ -1,6 +1,6 @@
 # Plan: close the MX-exclusion gaps in the trucking warmup
 
-**Status:** Fix 1 + Fix 3 SHIPPED 2026-06-20 (commit `9eeed47`); Fix 2 deferred.
+**Status:** ALL THREE FIXES SHIPPED 2026-06-20 (Fix 1+3 `9eeed47`, Fix 2 `bc93d93`).
 **Owner context:** warmup day 17; big operators (Google/Microsoft/Proofpoint/
 Mimecast/Barracuda/Cisco/Broadcom) are EXCLUDED until day 30, then re-introduced
 via `mx_daily_caps()`. This plan fixes three holes that let throttling/consumer
@@ -16,8 +16,12 @@ MX operators through during that window.
 - **Fix 3 (DONE):** `infra/cron/pw-mx-tag` installed to `/etc/cron.d/` (05:45 UTC
   daily, `--only-unsent --limit-domains 20000`). Verified: a 200-domain test run
   tagged 216 domains; idempotent/bounded.
-- **Fix 2 (DEFERRED):** bounding the NULL bucket cap — the cron will drain the
-  backlog, revisit if NULL stays large.
+- **Fix 2 (DONE):** `select_sendable_carriers()` now bounds the untagged (NULL
+  mx_provider) bucket with a single shared `untagged_cap` (env
+  `MAIN_UNTAGGED_MX_CAP`, default `max(quota, 200)` = no-starve / no behavior
+  change today). Only ~3,035 distinct verified-sendable untagged domains remain,
+  so pw-mx-tag drains them in its first run; tighten the cap to a fraction of
+  quota afterward to prefer the tagged long tail. Commit `bc93d93`.
 
 ---