DNS propagation versus a configuration error is a two-resolver test, not a feeling. If 8.8.8.8 and 1.1.1.1 disagree on one exclusive MX set, you are waiting on cache. If both still show Google beside MailerZ, leftover MX is published configuration. If you edited the registrar while Cloudflare owns NS, the clock never started. Missing aliases are not TTL. Do not toggle MX while you classify.
Quick answer for dns propagation vs configuration error
IETF RFC 1035 — Domain names lets resolvers cache MX for the old TTL. That wait is propagation. IETF RFC 5321 — Simple Mail Transfer Protocol then delivers to whatever set a given sender still believes. A registrar banner that says “updated” is not a resolver pair. How long the wait lasts is already on MX record propagation. This page is how you know you are in that wait at all.
Configuration errors that people call propagation: leftover Google, Microsoft, registrar, or null MX still published; wrong DNS panel; verification TXT missing; printed local-part never created; dual MX “for safety.” Time will not delete a record you still publish. Leftover MX is the usual config miss.
MailerZ Free is one domain, ten aliases, 14-day store, one seat, send-as disabled, SMTP and API disabled. Solo is $40 per year. Starter is $8 or $80. Business is $19 or $190. Agency is $39 or $390. Unlimited is $99/month or $990/year. Confirm pricing. None of those cards flush Google Public DNS or forgive a second hostname you left in the zone.
Related: change MX without losing email, authoritative NS, DNS looks correct, MX to inbox. Use those after you name the class.
Tools: MX lookup and leftover MX from two public resolvers. Your laptop cache is not one of them.
The user problem and the decision criteria
Founders wait 48 hours because a blog said so, while both public views still list ASPMX. They are not waiting. They are misconfigured. Other founders toggle MX every ten minutes because one resolver is still on the old exclusive set. They turn a one-hour cache into a week of oscillation. The decision is wait versus delete versus change panels versus create an alias. Mixing those four into “propagation” is the outage.
| What two public resolvers show | Class | Action |
|---|---|---|
| Different exclusive sets, no leftover brand | Propagation | Wait the old TTL. Do not toggle. |
| Both list MailerZ and Google | Configuration | Delete leftovers. Not a wait. |
| Neither matches the panel you edited | Wrong NS / panel | Edit the authoritative host. Clock not started. |
| Both exclusive MailerZ, one alias missing | Route config | Create the alias. Check hold. |
| You changed MX five times today | Oscillation | Publish once. Sit. |
| Only your laptop disagrees | Local cache | Public views win. Flushing your PC is not the internet. |
Agencies write the class in the ticket title: “propagation 8.8 vs 1.1” or “leftover Google both views.” That sentence prevents a contractor from adding the old host “until propagation finishes.” Dual MX is never how a wait ends.
Self-send during either class lies. Gmail may never leave Google. Classify DNS first, then probe from a mailbox you do not own. Self-send tests.
Technical mail flow: cache versus the zone you published
Authoritative nameservers answer the current MX set. That answer is configuration. Recursive resolvers copy it until TTL ends. A sending MTA asks its resolver, not your dashboard. During a real wait, some senders still hit the old exclusive set and some hit the new exclusive set. Both can be true. That is dns propagation vs configuration error as cache, not as leftover.
If the authoritative zone still contains two operators, every new lookup can return both. Caches will keep returning both after you “wait.” That is why leftover MX never expires. Time is not a delete key.
MailerZ accepts a named alias, stores required content, returns 250, rewrites envelope MAIL FROM with SRS, and leaves Header From intact. See email forwarding. None of that runs if the sender still connected to leftover Google. History will be empty. Empty history during a true cache wait is possible for senders on the old view. Empty history after both public views show exclusive MailerZ plus leftovers gone is a sender or alias problem, not TTL.
Wrong panel: public NS points at Cloudflare and you edited GoDaddy. Authoritative answers never changed. Resolvers agree with each other and disagree with your screenshot. That is configuration. The wait has not begun.
Free holds unknown recipients. That is product policy, not DNS. Paid may forward unknown after you choose it. Creating billing@ is configuration. Waiting overnight will not invent the row.
SPF, DKIM, and DMARC do not route inbound. Waiting for SPF “to propagate” so hello@ arrives is a class error. Publish sending records when you send. Do not block the inbound tell on them.
The tell, in order
Read public NS
If it is not the account you have open, stop. You are not in a TTL window.
Query two public resolvers for MX
Hostnames and preferences. Screenshot both. Include leftover brand names if present.
Compare to the authoritative panel
Panel should already be exclusive MailerZ if you finished the cut. If the panel still has Google, that is config even if one resolver looks clean.
Apply the table
Disagree on exclusive sets: wait. Agree on two companies: delete. Agree on exclusive MailerZ: probe aliases. Do not invent a fourth class called “give it 48 hours.”
Only then send a unique probe
Third mailbox. After views agree on exclusive MX. Self-send off.
Lowering TTL after you already published the new set does not rewind caches that hold the old exclusive set. It helps the next change. The wait is the TTL that existed on the set you replaced. If you never lowered TTL yesterday, do not pretend today’s 60-second value applies to yesterday’s 86400.
Corporate resolvers can ignore TTL. One office that never updates is not the internet. Prove public views first. Then tell that office their resolver is sticky. Do not dual-publish for them.
Step-by-step: classify, then wait or fix
Freeze edits while you classify
No MX toggle for ten minutes. Oscillation destroys the tell.
NS, then two MX lookups
Write UTC. Use leftover MX.
If leftover names appear in both views, delete them
Same hour. Demoting preference is not deletion. MX priority.
If views disagree on exclusive sets, wait the old TTL
Write the check time. Do not add Google back “to help the slow resolver.”
When views agree exclusive, probe inbound
Named aliases. Unique subjects. Classify hop if still missing. Rejection diagnosis.
Announce after the probe
Not after the registrar badge.
Rollback, if you truly must, is exclusive old screenshot, not two operators. MX rollback. Rolling back because you mislabeled leftover MX as TTL wastes the day.
Send-as is not this tell. Free has no send-as. A 535 during a wait is a pair problem. AUTH failed.
Failure modes people call propagation
Leftover MX forever. Proof: both resolvers list a second company. Fix: delete. Time will not.
Wrong panel. Proof: public NS ≠ your login. Fix: real host. Clock never started.
Missing TXT. Proof: domain not verified. Fix: publish dashboard TXT. Waiting does not verify.
Missing alias. Proof: exclusive MX, hold or 550. Fix: create the name.
Toggle storm. Proof: five MX edits. Fix: one exclusive set. Sit one old TTL.
Laptop flush theater. Proof: only your PC changed. Fix: public views.
Self-send. Proof: Gmail short-circuit. Fix: third mailbox.
SPF wait. Proof: inbound ticket blocked on SPF lookup. Fix: MX path first.
Registrar upsell. Proof: parking MX after a phone call. Fix: delete. Lock editors.
Null MX leftover. Proof: refuse-all still published. Fix: delete with the real set.
Sticky office resolver. Proof: public exclusive, LAN old. Fix: that network. Not dual MX.
421 at destination after exclusive MX. Proof: hop two 4xx. Fix: wait dest. Not DNS class. SMTP 421.
MailerZ workflow and the product boundary
MailerZ shows the MX to publish. It does not flush 8.8.8.8. It is not Workspace, not IMAP or POP, not an open relay, not SOC 2. Security. Docs.
Header From stays intact. Envelope SRS only. Seats operate the dashboard. They do not expire caches.
Store 14 days on Free, 90 on paid. Useful after MailerZ accepted. Useless for leftover Google accepts. Empty hop during a real wait can be the old view. Empty hop after exclusive public MX is a different owner.
Compare: Cloudflare Email Routing, ImprovMX. Their old TTL is still their old TTL.
Unauthorized send-as is 550. Not a propagation story.
Cost of a correct class versus a 48-hour myth
Two lookups are free. A courtesy leftover is expensive. Solo at $40 per year does not buy a shorter TTL. Starter $8/$80, Business $19/$190, Agency $39/$390 scale domains and send ceilings. Annual paid except Solo includes two months free versus monthly. Confirm pricing.
Do not buy Agency to propagate faster. Do not buy Google Workspace — product overview because one resolver is late. Buy a suite for Calendar, Drive, and hosted stores.
Staff time: one person owns the class. Slack “is it propagated” every ten minutes is how toggles start. Put UTC check time in the channel.
Premium DNS answers faster. It does not erase caches that already hold the old exclusive set. Anycast is not a leftover delete.
Tools on tools are enough. You do not need a paid monitor to tell leftover from TTL.
Legal does not get a seven-year archive from a wait. Say the store window out loud.
Field notes
Ticket titles: “leftover Google both views” versus “8.8 MailerZ / 1.1 old exclusive.” Those two titles are different runbooks. Mixing them is the week-long outage.
If the client insists “DNS always takes 48 hours,” show both lookups. If they match on two companies, you have a teaching moment and a delete. If they disagree on exclusive sets, show the old TTL you screenshotted before the cut. If you forgot the TTL, assume the worse common value and still query instead of guessing forever.
Phone registrar apps during a wait republish old MX. Treat the zone like production. One owner. Courtesy leftovers are how dns propagation vs configuration error never ends.
After views agree, do not skip the inbound probe. Exclusive MX with no alias is a config miss that looks like “still propagating” to a founder who only tests billing@.
IPv6-only resolvers and IPv4 senders can theoretically see different glue if NS is broken. That is an NS ticket. Do not dual-publish MX to paper over broken nameservers.
Write the customer sentence before the cut: “Inbound may split for up to N minutes while caches expire. We will not run two mail hosts.” That sentence prevents the leftover courtesy request.
Night on-call: if both public views already exclusive MailerZ and mail is missing, you are on the MX-to-inbox path, not on TTL. Do not add Google at 11 p.m.
Friday 4 p.m. cuts with a 86400 TTL are Monday problems if you call the wait “broken config” and toggle. Say the clock in the customer email before the cut. If you cannot wait a day, lower TTL a day earlier next time. Lowering it an hour before the cut does almost nothing.
Two founders sharing a zone without a screenshot will argue class from memory. Date the lookups. The tell is a file, not a Slack emoji.
If hop history is empty and both views exclusive MailerZ, you do not have a propagation ticket. You have a sender who never sent, a typo, or a subject you are searching wrong. Reproduce with a nonce. Troubleshooting.
Catch-all hold filling during a wait is not proof MX is exclusive. Unknown names can arrive on MailerZ while leftover Google still eats hello@. Classify MX first. Then read hold. Mixing those is how people enable catch-all forward to “fix propagation.”
IPv6 glue errors are NS configuration. Waiting 48 hours will not repair broken nameservers. Fix NS, then re-run the two MX lookups. Do not paper NS with a second mail host.
Write the on-call rule in the runbook: no MX edits unless leftover names appear in both public views or NS is wrong. A single slow resolver is not leftover. A leftover in both views is not a slow resolver.
If you inherited a zone with no pre-cut TTL screenshot, query now and treat the higher of the two public TTLs as the wait if views already disagree. If views already agree on leftovers, you are not waiting at all. Delete. Guessing “48 hours” is how leftover Google survives a long weekend.
Document dest 421 separately. A hop-two 421 after exclusive agreed MX is not dns propagation vs configuration error. It is a destination retry. Mixing those classes is how people add Google MX during a Gmail greylist.
Teach the registrar badge last. Teach the two lookups first. A green badge over a leftover host is still leftover. A grey badge over exclusive public MX is still exclusive. The internet does not read the badge.
Worked scenarios for cache versus leftover
Two public resolvers disagree: one shows only MailerZ, one still shows the old host. You are in a cache window. Wait the TTL you actually read. Do not toggle records while you wait. Do not call it leftover yet.
Both resolvers show Google plus MailerZ. That is leftover MX. Configuration. Delete. The clock is not the problem. Dual owners are.
You edited a panel whose NS are not public. Both resolvers still show the old set and will until you change registrar NS or edit the live zone. The clock never started. Ghost panel is configuration, not propagation.
Aliases are missing. HOLD has the customer. That is not DNS propagation. That is a map miss. Do not wait out a TTL for a name you never created.
A checker is green and strangers still miss. Checkers are not two resolvers. Query resolvers. Then probe from a third mailbox. Self-send can hide leftovers during a real wait.
Practice and anti-patterns when you classify the wait
Practice: two public resolvers, then leftover tool, then probe. Anti-pattern: refresh the console for an hour.
Practice: read TTL. Anti-pattern: folklore twenty-four hours.
Practice: delete leftovers when both views agree on two companies. Anti-pattern: waiting for leftover to 'propagate away.' It will not.
Practice: confirm NS first. Anti-pattern: a beautiful unused zone.
Practice: one class per ticket. Anti-pattern: toggle MX while classifying cache.
Practice: foreign probe after views agree. Anti-pattern: Gmail self-send as the all-clear.
Operator closeout after you named the tell
Closeout says cache window, leftover configuration, ghost panel, or map miss. One sentence. Then the listings, TTL read, action, and next probe time.
If you waited, the later listings belong in the packet. If you deleted leftovers, before and after belong.
Agencies file per zone. 'It's propagating' as a fleet status is how the tenth zone stays leftover.
Link /tools and /troubleshooting. RFC 1035 is what resolvers implement.
No inboxing percent at the end of a wait. Agreement on MX is not a tab.
Edge cases in propagation stories
NS TTL versus MX TTL. You changed NS. The long clock is NS. MX TTL does not save an empty new zone.
DNSSEC SERVFAIL looks like 'not propagated.' Classify signatures before you wait a day.
Anycast resolvers that still hold the old set after others moved. Two views are the minimum. A third view helps if they split.
IaC apply that re-adds leftover MX every morning. That is configuration on a loop. Grep the repo. Not a cache.
Website wizard republish overnight. Morning disagreement after last night's exclusive is configuration again.
IPv6 resolver path versus IPv4 seeing different zones. Rare, but read both if the ticket smells like split views.
People who may not say it's just propagating
Support may say wait only after two resolvers disagree on one exclusive set. If both show two companies, they must say leftover, not wait.
DNS owners confirm NS. Website owners confess to wizards. Engineering greps IaC. Three different leftover sources.
Customers can wait a cache window. They should not wait a leftover. Teach the difference in one sentence.
Agencies forbid 'propagation' on tickets that lack two listings. No listings, no wait story.
Proof packet for the tell
Two resolver listings with timestamps, NS listing, TTL you read, class, action, leftover tool, probe after agreement.
If class is cache, no MX toggle in the window. If class is leftover, delete proof. If ghost, live NS proof.
Self-send not the all-clear. Third mailbox after views agree.
Filename the zone and UTC. 'Still propagating' without listings fails the packet.
Author: MailerZ editorial, Secuno LLC. Review when resolver UIs, builder wizards, or pricing change.
Buying a mailbox seat does not expire a cache and does not delete a leftover host. If both resolvers already agree on two companies, wait is the wrong medicine. Delete the extra exchanger, then probe from a mailbox you do not own.
If public NS still name a different panel than the one you edited, stop calling it propagation. Point the registrar at the zone you actually changed, or edit the live zone. The clock starts only after the internet can see the exclusive MailerZ set you meant to publish.
FAQ
How do you tell dns propagation vs configuration error?
Query two public resolvers. If they disagree on one exclusive MX set, you are in a cache window. If both still list Google plus MailerZ, leftover MX is configuration. If public NS does not match the panel you edited, the clock never started. Missing aliases and hold are not propagation.
Does this require a new mailbox?
No. The tell is DNS views versus the published zone. Destinations stay Gmail or Outlook. MailerZ is not IMAP. Buying a seat does not expire a cache or delete a leftover host.
Will it work with Gmail or Outlook?
Yes as destinations after exclusive MailerZ MX. Gmail as a sender may still use a cached MX during a real wait. Self-send can hide leftovers. Probe from a mailbox you do not own after views agree.
What DNS records are involved?
NS so you know the real panel, MX and its TTL, leftover host names, verification TXT. SPF does not decide inbound. A green checker is not a resolver pair.
What should I test before production?
Two public MX lookups, leftover tool clean, then a unique inbound probe. If resolvers disagree, wait the old TTL. If they agree on two companies, delete. Do not toggle while you classify.
Key takeaways
- Two public resolvers are the tell.
- Disagree on exclusive MX is propagation. Wait the old TTL.
- Both views listing two companies is leftover MX. Delete.
- Wrong NS means the clock never started.
- Missing aliases are not TTL.
- Hold is not propagation.
- Registrar banners are not proof.
- Do not toggle MX while you classify.
- Dual MX is never how a wait ends.
- Laptop flush is not the internet.
- Self-send can hide leftovers.
- SPF wait is the wrong class for inbound.
- MailerZ does not flush public DNS.
- MailerZ is not IMAP, not a suite, not SOC 2.
Conclusion and next action
DNS propagation versus a configuration error is named by two lookups and the panel that matches public NS. Time deletes caches. Time does not delete leftovers, wrong panels, or missing aliases.
Next action: screenshot NS and two MX views. If leftovers, delete. If exclusive sets disagree, wait the old TTL and do not toggle. If exclusive and agreed, probe each named alias from a third mailbox. If still missing, you left this page for hop diagnosis.
Keep the two screenshots next to the alias list. The next “is it propagated” should take two minutes.
Ready to classify before you wait
Start free, create aliases, publish exclusive MX.
Free receives one domain. Sign in if it is already there.
Review quarterly, or sooner if dashboard MX hostnames change. Author: MailerZ editorial, Secuno LLC.