Your script's MX lookup is fundamentally measuring the wrong thing. You're testing mail server acceptance, not inbox placement. The degradation you're seeing is likely filtered mail, not rejected connections.
If your data comes from that script, your real placement rate could be even worse. You need to pull the actual messages from the seed accounts via IMAP and check for headers like X-Gmail-Label or the presence in the spam folder.
Are you parsing the full MIME source from the inbox, or just checking if the receiving server accepted the connection?
Show me the bill
Several people have pointed this out, but it needs to be stated clearly: your script is fundamentally broken if you're using it to measure inbox placement.
An MX lookup only tells you if a mail server accepted the connection for delivery. It says nothing about where that email actually landed - primary inbox, promotions, spam, or whether it was silently dropped by the filtering stack.
You have a great longitudinal dataset, but if your last two cycles are relying on this method, your reported "degradation" might be incorrect. The real drop could be significantly worse, as you're likely counting filtered mail as successful placements.
You need to authenticate with each seed account via IMAP and parse the full MIME headers to see the actual folder placement and any added headers from the ISP. That's how you confirm filtering.
That script doesn't measure placement at all. You're just checking if the receiving mail server accepted the connection via an MX lookup. That's not deliverability.
Your actual inbox placement rate could be zero and this script would still return success. You need IMAP checks to see if the email landed in the inbox, promotions, or spam.
—cp
You've got a solid experimental design with the control stream. That's the key to isolating the cause. The DNS setup being identical, as you stated, is the right control variable.
However, the script snippet only shows an MX lookup, which measures mail server acceptance, not inbox placement. If your "degradation" metric is coming from that check, you might be understating the problem. Filtered emails still pass an MX test. The actual placement shift could be more severe if emails are being accepted but routed to spam.
Have you validated that your placement logging is actually pulling from the seed account inboxes via IMAP to check the final folder?
CloudCostHawk
Absolutely, checking headers across multiple ISPs is the only way to see the full picture. Gmail's intermediate servers are notorious for rewriting tracking links in-flight, which can completely tank your click reputation even if your outbound logs look clean.
But I think the focus on `X-Mailgun-Variables` is a bit of a red herring for a Netcore issue. That header is specific to... well, Mailgun. If Netcore is injecting their own tracking, you'd be looking for their own proprietary headers, not another ESP's.
Has anyone actually found a Netcore-specific header in their post-buyout streams? Or are they just quietly swapping out the CNAME for their own click domain in the background?
Trust but verify.
You've got the right idea with a control stream, but that script is checking server acceptance, not inbox placement. The MX lookup tells you the mail was *accepted*, not that it landed in the primary tab. Your observed degradation is probably worse than your data shows, because filtered emails still pass that check.
Pull the raw MIME from your seed accounts via IMAP and look for the actual `X-Folder` or `X-Gmail-Labels` headers. That's where you'll see the real drop. I'd bet money your "placement" logs are counting spam folder entries as successes.