What is working now, and what failed before

Incident history

Every documented incident, with its impact, its cause in plain words and what we changed so it does not happen again.

1. Oct 9, 2026WebsiteResolved on Oct 9, 2026

   A new version of the site could not be published

   Impact

   The afternoon release failed and that day's changes arrived late. No data or accounts were lost.

   Cause

   A page of the English and Portuguese site was missing translated texts that the Spanish version already had, and the site build stopped there.

   Fix

   The texts were added; since then a missing translation shows the Spanish text instead of stopping the release, and the full site is built before publishing.

2. Oct 9, 2026Nightly updateResolved on Oct 9, 2026

   The nightly update arrived about five hours late

   Impact

   Data for the 8 October session became available about five hours later than usual, before the regular session opened.

   Cause

   The download of the day's data failed twice because of temporary errors at the source service, and the job stopped without retrying.

   Fix

   The download now retries on its own with growing waits before it gives up. Since 11 October each night's result is published on this page.

3. Oct 5, 2026Nightly updateResolved on Oct 9, 2026

   Four pages stayed on 5 October data

   Impact

   Between 5 and 9 October, Seasonality, Market Profile, Macro data reaction and the expected range showed data through 5 October while the rest of the platform updated every night. Nothing flagged the delay.

   Cause

   The regular-session minute files behind those pages were rebuilt by hand and were not part of the nightly update.

   Fix

   Since 9 October the nightly update rebuilds them every night, only when the source is complete. The portal footer shows the session the data runs through and warns when it falls behind.

4. Sep 27, 2026APIResolved on Sep 28, 2026

   The API did not respond for about 17 hours

   Impact

   From the evening of 27 September to the morning of the 28th (New York time) the portal pages that request data did not load. The public site stayed available.

   Cause

   After a change in server capacity the service stopped answering requests. The automatic check only looked at whether the server was on, not whether it answered, so it did not restart it. The exact reason for the initial failure could not be determined.

   Fix

   The check now makes a real request and restarts the service when it does not answer. A second, independent check every 10 minutes reports the real state, and every hour the response time behind this page is measured.