This is happening to me as well after upgrading from 3.04 to 3.17 on a Framework 13 12th-gen.
To try to understand the problem a bit better, I started paying closer attention to temperature sensor readings, and it looks like the short bursts of speed on the fan are correlated with sudden spikes in the readings coming from ACPI temperature sensor 1.
I’ve added a screenshot below that I took while running stress to keep the CPU hot, showing that sensor inexplicably dropping to zero briefly. It’s capturing a different scenario than the one we’re complaining about, but it’s relevant because it confirms that the sensor readings aren’t reliable, and the unreliability has an impact on fan speed, despite the presence of other sensors that can be used to implement a more correct fan curve.
I don’t always consistently bother to report problems upstream. The fact that I’m bothering to post this message should demonstrate how distracting this problem is - it was enough to make me stop what I was doing and share this report instead. I’ve recommended Framework laptops to many of my coworkers, and anyone else who might be interested. I’ve owned mine for over two years without this problem, and if this is the new status quo, it will seriously affect how most users feel about the purchase.
I know firmware troubleshooting, development, and releases must be a royal pain to manage, but I think this problem should be handled with urgency. Even if it’s not possible to quickly find and fix the problem with that sensor, there are enough other sensors available to help control the fan that the firmware can afford to ignore this one. CC @Matt_Hartley, because of this message.