You're on the right track with the placement audit, but I'd argue it's premature as step one. The raw header analysis from a few seed sends is a faster, free diagnostic that can confirm or rule out the IP pool change theory in under an hour. If you see a new sending IP in the last external `Received` header, you've found a probable cause and can take that evidence directly to your ESP. An audit will tell you you're in Promotions, but the header tells you *why* that might have happened suddenly. Do the free check first, then use the audit for granular placement mapping if the headers don't show a smoking gun.
Mike
Totally agree on the test-first approach with a seed list. It's like checking the temperature before you jump in the pool.
Your point about X-Gm-Message-State is a pro move. For anyone else trying this, Gmail's Postmaster Tools API can be a bit of a bear to set up initially, but that granular placement data is gold once you get it flowing. A simpler hack I've used is to just have those seed accounts auto-forward a copy of the received message to a parser I set up in Make - the headers are all preserved, and you can spot "category: promotions" without needing full API access.
But you've hit the core tension: if even your best content lands in tabs for your most engaged seeds, you're not in a re-engagement problem anymore, you're in a hard reputation reset. That changes the whole game plan.