TL;DR: A quarantined phishing mail is not blocked. It is delivered to the spam folder, synced to every device, indexed for search and reachable through forwarding, shared mailboxes and mobile previews. The difference between
p=quarantineandp=rejectis the difference between "the message exists" and "the message does not exist".
The sentence that sounds like protection
"DMARC is enforced, policy is quarantine."
It reads like a closed door. It is a filing instruction. The receiver accepts the message, evaluates it, decides it is suspicious, and puts it in a folder. Then it does what mail systems do with every folder:
- it syncs it to the laptop, the phone and the tablet;
- it indexes it for search;
- it counts it in the badge on the app icon;
- it includes it in shared-mailbox views;
- it hands it to every rule that does not explicitly exclude spam.
The message has a size, a body, a date and a sender. It exists.
What actually happens at SMTP
Under quarantine, the server conversation succeeds:
S: 250 2.1.0 OK MAIL FROM:<bounce@relay.tld>
S: 250 2.1.5 OK RCPT TO:<inbox@example.tld>
S: 354 Go ahead
S: 250 2.0.0 OK queued as 4XYZ
The filtering happens after acceptance. That is the whole architecture: SMTP accepts, the policy engine evaluates, the message is filed.
Under reject, the same conversation ends differently:
S: 354 Go ahead
S: 550 5.7.1 Message rejected due to DMARC policy
No message id. No body. No spam folder. No sync. The sender gets a bounce, which is the point: a legitimate sender who is misconfigured learns about it immediately instead of silently losing mail.
Three ways a quarantined message comes back
Sync. The spam folder is not a server-side quarantine that requires an administrator. It is a folder, and folders replicate. Every device now holds a copy of the phishing mail.
Search. Mailbox search indexes spam. A user looking for an invoice number can find the forged invoice — and the context around it.
Forwarding and shared mailboxes. A rule that forwards "all mail" or a shared inbox that a team monitors does not necessarily respect the spam classification. The message re-enters the normal flow without anyone approving it.
There is a fourth, softer path: curiosity. A spam counter on a phone is a notification. It says: something arrived, and it was interesting enough to be filtered.
The fear that keeps teams at quarantine
Three objections, all reasonable, all answerable with data:
"Our invoice mails might break." Check the reports. If no legitimate sender fails both SPF and DKIM, nothing breaks. If one does, fix that sender first — that is a finding, not a reason to keep the door open.
"Forwarding will break." Forwarders keep DKIM, and DMARC passes on DKIM. This is the most common false fear.
"We might lose a customer mail." Under quarantine you keep it — in a folder the customer does not know about. Under reject the sender gets a bounce and resends. Losing a mail silently is the worse failure mode.
Findings
| Under quarantine | Under reject |
|---|---|
| message delivered, filed as spam | message refused at SMTP |
| exists on every synced device | exists nowhere |
| searchable | not indexed |
| can be forwarded by rules | never enters a rule |
| visible in mobile badges | no notification |
| sender never learns | sender gets a bounce |
The decision rule
no failures -> reject now
all failures already quarantined -> reject now
a known legitimate sender fails -> fix that sender first
no rua reporting at all -> stay at quarantine, add rua, wait two weeks
The last line matters: moving to reject without reports is not a hardening step, it is a bet.
Closing
Quarantine is a useful intermediate state. It is not a destination. A domain that has been at p=quarantine for two years has not been protected for two years — it has been collecting phishing mail in a folder that everyone can read.
The question is not whether the policy is "enforced". The question is whether the message still exists when a human decides to trust it.
Once an attacker gains access, they can perform the following actions: