After a recent OTA firmware update, some devices are reporting Time.now() values a couple minutes in the future from when they were received which we've never noticed before. This device tells us that it published this event 90 seconds in the future from when we actually published it.
When we publish a message, we write "created_at": Time.now() directly into the call to Particle.publish with json data.
You can see here that the created_at field is 1777482281 or 4/29/26 at 10:04:41 am in PDT. The published at time (which is accurate) indicates 4/29/26 at 10:03:07 am.
The Tracker has the ability to load the Device OS Time class from the AM1805 RTC/watchdog chip at boot. My guess is that the chip time is out of sync with cloud time. It should eventually reset after cloud connection and time sync, but there will be a period of time where Time.isValid() is true but it's using the stored value. Does it ever go back to correct, or is it staying at the wrong time? Which Device OS version are you using?
About 1% of our devices are showing future created timestamps, associated with them rejecting application configuration on startup (not a device error, but an application error that made them want to reset immediately after contacting our services). It looks like each time this happened, a little error would creep in somehow that would get cleared periodically. Once we fixed the application problem causing immediate reset, it looks like the clock drift stabilized.
That seems plausible. The time sync happens shortly after connecting to the cloud, so if you don't get to that point, it would never synchronize. It's possible to publish an event before the time sync happens.
