docs: record trucking per-MX throttling fix + the 702k-google smoking gun

docs/vertical-lead-source-analysis.md

@@ -101,3 +101,22 @@ Unlocked the full verified institutional pool for broad offers:
   dated path that rotates ~monthly. If the download 404s, re-fetch the dataset's
   current downloadURL from https://data.cms.gov/data.json. Consider switching to
   the dataset's stable data-api endpoint.
+
+## Update 2026-06-14: trucking/main pool per-MX throttling (deliverability fix)
+
+The persistent main-pool 54% delivery + Gmail/Outlook block storm (Jun 13-14)
+root cause, now PROVEN by MX-tagging the carrier pool:
+- **702,214 carriers on Google + 135,129 on Microsoft** -- the warmup was
+  hammering exactly the two operators blocking us (no per-MX throttle on trucking,
+  only HC had it).
+- Fix: migration 097 (mx_provider) + mx_tag_carriers.py (concurrent MX resolve,
+  bulk temp-table-join write -- 1.24M/1.49M carriers tagged). build_trucking_
+  campaigns now EXCLUDES Google/Microsoft/Proofpoint/etc. until warmup day 30
+  (reputation recovery), per-MX caps thereafter. Untagged carriers pass (most are
+  now tagged).
+- Effect on the MCS-150 overdue pool: 496,743 sendable -> 230,135 after excluding
+  263,515 Google/MS carriers. Plenty of long-tail volume (yahoo/comcast/charter/
+  centurylink/windstream/earthlink/...) to warm on safely while reputation recovers.
+- MAINTENANCE: re-run mx_tag_carriers.py periodically (or add to the trucking
+  cron precursor) to tag newly-added carriers; flip MAIN_BIG_MX_EXCLUDE_UNTIL_DAY
+  or MAIN_SKIP_BIG_MX=0 once Postmaster Tools shows recovered reputation.