Production incident: Nylas v3 API timeouts at ~90s across multiple grants on 2026-06-25

Question and Answers
1

Hi Nylas team,

We tried submitting this through the Nylas support form, but the form failed with an internal error, so posting here to get this routed.

We are seeing a production incident today, 2026-06-25, with elevated Nylas v3 API latency/timeouts across multiple Gmail-backed grants. This is affecting both outbound sends and message-list reconciliation.

Environment:

  • Nylas API v3
  • Node/TypeScript SDK: nylas@8.0.5
  • Host: api.us.nylas.com
  • Main error: “Nylas SDK timed out before receiving a response from the server.”
  • Time window: roughly 2026-06-25T16:30Z through at least 2026-06-25T17:20Z
  • Public status page currently shows no incident for Jun 25.

Affected endpoints:

  • GET /v3/grants/{grant_id}/messages
  • POST /v3/grants/{grant_id}/messages/send

Observed behavior:

  • Requests are timing out at almost exactly 90 seconds.
  • We saw over 100 matching Nylas/email failure logs today.
  • A reconciliation sweep at 17:00 UTC produced 39 Nylas SDK timeout logs across 25 orgs/grants.
  • One representative GET /messages trace lasted ~89,999ms against api.us.nylas.com before failing.
  • Database/internal work around the same trace completed quickly, so this looks isolated to the Nylas/provider API call.

Most important issue:
For POST /messages/send, some sends return timeout/provider/internal errors, but the message may still later appear in our local cache via Nylas webhook/message ingestion. That means retrying blindly risks duplicate customer emails.

Representative trace evidence from our side:

  • 2026-06-25T17:00:01Z-17:01:32Z

    • GET /v3/grants/{grant_id}/messages
    • Duration ~90.8s
    • Error: “Nylas SDK timed out before receiving a response from the server.”
    • Stack includes nylas@8.0.5 APIClient.sendRequest → Messages.fetchList → Messages.listIterator

  • 2026-06-25T16:48Z-17:13Z

    • Multiple POST /messages/send attempts
    • Error: “Email send failed: Nylas send failed: Nylas SDK timed out before receiving a response from the server.”
    • At least one timed-out send appears to have later succeeded / appeared through message ingestion.

Questions:

  1. Is Nylas seeing elevated latency/timeouts today for api.us.nylas.com, especially Gmail-backed grants?
  2. Is the ~90-second timeout enforced server-side by Nylas for all requests, regardless of SDK/client timeout?
  3. For POST /messages/send, can the provider still accept/send the email after Nylas returns a timeout/internal/provider error to the client?
  4. What is the recommended reconciliation process for ambiguous send outcomes before retrying?
    • Wait for message.created/message.updated/message.send_success?
    • Query Sent folder via /messages?
    • How long should we wait before deciding a timed-out send did not send?
    • Are Idempotency-Key retries safe if the original request timed out before any response?
  5. Should we reduce GET /messages page size below 50 or add specific select fields during reconciliation sweeps?

We have affected grant IDs, exact timestamps, and internal traces available, but I’m not posting grant IDs publicly. Please let us know the safest way to share them with Nylas support/engineering.

This is customer-impacting because blind retry risks duplicate customer emails, while not retrying risks missed sends.

Thanks,
Monk Engineering