Checking an email list through proxies: SMTP verification without burning your IP

Before a mailing goes out, somebody checks the list: MX lookup, a connection to the mail server, a RCPT TO for each address to see whether the mailbox exists. Do that for a hundred thousand addresses from one IP and Gmail, Outlook and Yandex stop answering you by lunchtime. This page counts what the check costs in addresses and why the free lists on this site are the wrong tool for it.

What the mail servers see

A verification probe opens a connection, says HELO, names a sender, asks about a recipient and disconnects without sending. Mail servers treat a stream of those from one address as a dictionary attack. Gmail starts greylisting after a few hundred probes an hour, Outlook returns a temporary error and remembers the address for a day, Yandex answers everything with "exists" once it has decided you are scanning. That last one is the worst outcome: the check keeps running and the results are wrong.

Counting addresses

A safe rate is about 200 probes per address per hour, spread across mail providers, with a pause of ten to twenty seconds between probes to the same provider. A list of 100,000 addresses checked within one working day needs 12,500 probes an hour, which is 60 to 70 addresses working the whole day. Add half again for the ones that get a temporary block and rest, and you are at a hundred.

A pool of thousands of addresses turns that into a few probes per address per day, below anything a mail server counts. RemProxy keeps a page for mailing work with a thread calculator and a Python sample that runs the SMTP check through the pool: a pool for mailing checks. The page also says what the sample does not do: it verifies addresses, it does not send.

Why a free proxy fails this check before it starts

Public proxies are open relays somebody forgot to close, and the blocklists know each one of them. Spamhaus, Barracuda and the provider's own lists carry most addresses you will find on a free list, and a mail server checks the connecting IP against those lists before it says hello. The probe gets a 554 or a dropped connection, and your script records the mailbox as invalid. Run a free list through an SMTP check and you will "clean" real customers out of your base.

Free lists stay useful for the parts of the job that do not touch port 25: pulling MX records, checking a domain's website, testing your script on a hundred addresses.

Sending is a different job

Send from a warmed domain through an ESP or your own server with SPF, DKIM and a PTR record. Proxies belong in the verification step and in reading bounce pages, never in the sending step. A message that arrives through a proxy IP has no PTR, no reputation and no SPF match, and lands in spam or nowhere.

Before the first run

• Probes to one provider from one address: no more than 200 an hour, ten seconds apart.
• Temporary errors (4xx) rest the address for thirty minutes and retry the recipient later.
• A sample of a hundred known-good addresses checked first; if more than five come back invalid, the block is on your addresses and the mailboxes are fine.
• Results stored with the response code, so a 554 from a blocklist is never confused with a missing mailbox.
• The catch-all test: probe a random mailbox on each domain, and treat domains that accept everything as unverifiable rather than valid.