What this is starting to look like is that the power budget was validated against a Wi-Fi-only device, and the dual-function MT7921AUN (Wi-Fi + Bluetooth) was either swapped in very late or not fully validated under simultaneous operation.
A few things line up too neatly to ignore:
-
The MT7921AUN is a USB combo device; Bluetooth initialisation causes short but real current spikes.
-
With the current Rset value (~6.8k), the USB rail is effectively capped at ~1A.
-
Wi-Fi alone stays stable.
-
The moment Bluetooth is enabled, the device browns out and the USB controller resets.
-
Reducing Rset (or adding proper buffering) resolves the issue.
On top of that, the antenna mount PCB has extra connections beyond what was originally advertised, which strongly suggests the design pivoted late to support a combo card rather than a Wi-Fi-only module.
So to be clear:
This doesn’t look like a Linux driver problem, and it doesn’t look like a faulty MT7921AUN. It looks like the final hardware configuration cannot reliably power the card in its advertised dual-function mode.
If the expectation is that users must:
-
rework a 0402 resistor, or
-
accept Bluetooth being unusable,
then that needs to be stated plainly — because as shipped, the kit is sold as supporting Wi-Fi and Bluetooth, but only one of those works reliably without modification.
Happy to test further, but at this point the evidence points squarely at USB power delivery being under-spec’d for the chosen card, likely due to a late hardware change.