Fix it now
Exchange started an SMTP conversation with the next hop and it did not finish. Both codes are transient, so mail sits in Retry rather than bouncing – until the queue lifetime runs out and the whole backlog expires at once as 4.4.7.
Get-Queue | Where-Object {$_.Status -eq 'Retry'} | Format-List Identity,NextHopDomain,MessageCount,LastRetryTime,LastError
Test-NetConnection <next hop host> -Port 25
Set-SendConnector 'To Internet' -ProtocolLoggingLevel Verbose
- Force a retry with
Retry-Queue -Identity '<server>\<queue>' -Resubmit $true, then read the send protocol log for the session that follows. - Look at where the log stops. A session that dies after EHLO points at inspection; one that dies at or after STARTTLS points at TLS; one that dies part-way through message data points at packet size or an idle timeout.
- Check your own certificate if TLS is where it stops:
Get-ExchangeCertificate | Format-List Thumbprint,Subject,NotAfter,Services. An expired certificate bound to SMTP breaks outbound TLS. - Check how long you have:
Get-TransportService <servername> | Format-List MessageExpirationTimeout,TransientFailureRetryCount,TransientFailureRetryInterval. The documented default expiry is 2 days. - Put protocol logging back to its previous level when you are finished, because verbose logging fills disk quickly.
Test from the Exchange server itself. A workstation takes a different path and often a different firewall policy, so a successful test from your desk tells you nothing about the queue.
If the session now completes and the queue drains, you are done. If it keeps dying at the same point, the next section maps each stopping point to a cause.
Why it happens
SMTP is a conversation and every stage can fail differently. Exchange opens a TCP connection, reads the banner, sends EHLO, negotiates TLS if it is offered, then sends the envelope and the message body. A failure at any point leaves the message queued with a transient error. The value of protocol logging is that it shows you which line of the conversation was the last one, and that single fact usually identifies the cause without further testing.
Be honest about what the codes themselves tell you, which is very little. Microsoft publishes no meaning for the 441 or 471 prefixes; they are Exchange queue text rather than standard SMTP replies. The enhanced part is standard: RFC 3463 defines the 4.4.1 class as an outbound connection attempt that was not answered, because the remote system was busy or unable to take a call, and the 4.4.2 class as a connection that was established but could not complete the message transaction, because of a timeout or inadequate connection quality. That distinction – never answered against answered and then lost – is worth more than the prefix.
4.4.7 is the one Microsoft does publish, and it is the deadline the others are running against. The documented diagnostic is #550 4.4.7 QUEUE.Expired; message expired ##, described as the rejecting system considering the message too old. In Exchange Online, delivery is retried over 24 hours before that report is generated; on premises the equivalent is the transport service’s message expiration timeout, whose documented default is 2 days.
Something is inspecting SMTP in the path
You have this one if The session opens and stalls at or just after EHLO, and the destination works when tested from outside your network.
- Ask the network team whether SMTP inspection, fixup or an application-layer gateway is enabled for traffic from the Exchange server.
- Have it disabled for the Exchange server’s outbound path. Mail security belongs in a mail gateway, not in a packet inspection engine that rewrites the protocol.
- Retest with protocol logging on and confirm the conversation proceeds past EHLO.
- Check the same behaviour inbound if you also receive mail directly.
This is the most productive thing to check when many destinations fail at once but only from one network.
TLS negotiation is failing
You have this one if The log shows the session ending at or immediately after the TLS handshake, and the destination advertises STARTTLS.
- Check your own certificate with
Get-ExchangeCertificate– an expired certificate bound to SMTP breaks outbound TLS. - Confirm the protocol versions and cipher suites enabled on the Exchange server are still compatible with what the destination requires.
- If one partner is affected, check whether the send connector requires TLS for their domain and whether their certificate has changed.
- Do not solve this by turning TLS off. Fix the certificate or the protocol configuration.
Large packets are being dropped in the path
You have this one if Sessions to many destinations die part-way through message data, always around the same size, and small messages go through.
- Test whether large payloads survive the path between the Exchange server and the destination network.
- Take that evidence to the network team. The usual causes are a tunnel or VPN in the path and blocked path discovery messages.
- Do not adjust interface settings on the Exchange server as a first response; fix the path.
- Retest by sending a message of the size that previously failed.
The destination is throttling or refusing you
You have this one if The destination answers, returns a temporary refusal and closes, often as 421 4.4.2, and the pattern follows their busy periods or a burst of mail from you.
- Check whether you have sent an unusual volume to that destination. Bulk mail from a normal mail server is frequently rate-limited.
- Slow down. Exchange retries on its own schedule and most greylisting clears on the second or third attempt.
- If it persists, contact the destination’s postmaster with the timestamps and your sending IP address.
- Confirm your sending address has a valid reverse DNS record, since some destinations refuse sessions from addresses without one.
Messages are expiring before the path is fixed
You have this one if After some days in Retry, a batch of non-delivery reports arrives at once carrying 4.4.7.
- Read the current settings:
Get-TransportService <servername> | Format-List MessageExpirationTimeout,TransientFailureRetryCount,TransientFailureRetryInterval. The documented default expiry is 2 days. - If a genuine outage will outlast the timeout, extend it deliberately so mail is not lost, and set it back afterwards.
- Resubmit queued messages once the path is repaired rather than waiting for the retry interval.
- Tell affected senders which messages expired, so they resend rather than assume delivery.
Full reference
Reading where the session died
| Last thing in the protocol log | Where to look |
|---|---|
| Nothing at all, the connection never opened | Firewall, routing, or the destination is offline |
| Banner received, then silence after EHLO | SMTP inspection on a firewall rewriting or blocking the exchange |
| Session ends immediately after STARTTLS | TLS negotiation, an expired certificate, or an inspection device stripping TLS |
| Fails part-way through message data, repeatably | Large packets being dropped in the path |
| The destination returns a 421 and closes | The target is refusing new work: overloaded, throttling, or greylisting you |
| Works for small messages, fails for large ones | Path packet size, or a per-message timeout at the destination |
What each code in this group is actually backed by
| Code | Backed by | Meaning |
|---|---|---|
| 441 4.4.1 | RFC 3463 for the 4.4.1 class only | The outbound connection attempt was not answered – the remote system was busy or unable to take a call |
| 471 4.4.1 | Nothing published by Microsoft | Exchange queue text. Diagnose it from the protocol log, not from the number |
| 421 4.4.2 | RFC 3463 for the 4.4.2 class only | The connection was established but could not complete the transaction, from a timeout or poor connection quality |
| 550 4.4.7 | Microsoft Learn | QUEUE.Expired; message expired. The rejecting system considers the message too old |
It is worth being blunt about this: three of the four are not published by Microsoft, and any confident sentence you read about what 471 4.4.1 means came from somebody’s experience rather than from documentation. The protocol log is the evidence.
Turning on protocol logging without regretting it
- Note the current level so you can put it back:
Get-SendConnector | Format-List Name,ProtocolLoggingLevel - Set it to Verbose on the connector that carries the failing traffic only.
- Force one retry and let it fail once, rather than leaving logging on across a peak.
- Read the log under the transport service’s logging directory, find the session by destination and timestamp, and note the last command exchanged.
- Set the level back to None. Verbose protocol logging on a busy connector fills disk quickly, and a full disk on a transport server triggers back pressure.
What Microsoft documents for 4.4.7
- Test outbound mail flow with the Outbound SMTP Email test in the Remote Connectivity Analyzer.
- Verify that the SPF record is complete and includes every source that sends for the domain.
- Confirm with the domain registrar that the domain is active and has not expired.
- In a hybrid deployment, rerun the Hybrid Configuration Wizard, since on-premises address or firewall changes can leave the configuration stale.
- Bear in mind that in Exchange Online delivery is attempted repeatedly over 24 hours before the report is generated, so the sender learns late.
Extending the message expiration timeout during an outage is a deliberate stopgap, not a fix. It delays the moment senders learn their mail did not arrive, and a long queue on a transport server eventually becomes a disk problem. Put it back when the path is repaired.
Clearing a stuck queue safely
Resubmitting is safe and is usually what you want: Retry-Queue -Identity '<server>\<queue>' -Resubmit $true puts the messages back through categorisation and routing, which is how they pick up a corrected connector or a corrected DNS answer. Removing messages from a queue destroys them, with or without a non-delivery report depending on how you do it, so treat that as a last resort and record what you removed and for whom.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
441 4.4.1 |
Microsoft publishes no meaning for the 441 prefix. RFC 3463 defines the 4.4.1 class as an outbound connection attempt that was not answered because the remote system was busy or unable to take the call | RFC 3463 |
471 4.4.1 |
Neither the 471 prefix nor an Exchange-specific meaning is published by Microsoft. Diagnose it from the send protocol log rather than from the number | not published by the vendor |
421 4.4.2 |
Not published by Microsoft. RFC 3463 defines the 4.4.2 class as a connection that was established but could not complete the message transaction, from a timeout or inadequate connection quality | RFC 3463 |
550 4.4.7 |
QUEUE.Expired; message expired. The rejecting system considers the message too old; Exchange Online retries over 24 hours before generating it | Microsoft Learn |
Confirm the fix worked
Get-Queueshows the affected queue Active or empty rather than in Retry.- A verbose send protocol log for a new session shows a complete conversation ending in acceptance by the destination.
Test-NetConnection <next hop> -Port 25from the Exchange server succeeds consistently, not once.- Protocol logging has been set back to its previous level.
- No new 4.4.1 entries appear in queue LastError across a full business day.
Questions people ask about this
Do I need to buy anything to fix this?
No. Every cause here is a network path, a certificate or a configuration setting. There is no licence that makes an SMTP session complete.
What do 441 and 471 actually mean?
Microsoft publishes no meaning for either prefix. The 4.4.1 part is standard – a connection attempt that was not answered – and the rest has to come from the protocol log. Be sceptical of any confident explanation of 471 that does not cite a vendor page.
Should I just increase the queue lifetime?
Only as a deliberate stopgap during a known outage. The documented default is 2 days. Extending it hides the problem and delays the moment senders learn their mail did not arrive, so put it back once the path is repaired.
Why does the same destination work from my laptop?
Because your laptop takes a different path, and often a different firewall policy, from the Exchange server. Always test port 25 from the Exchange server itself.
Can I safely clear the stuck queue?
Resubmitting is safe and is usually what you want. Deleting messages from a queue destroys them, with or without a non-delivery report depending on how you delete, so treat that as a last resort and record what you removed.
