After upgrading my Comet PoE (RM1PE) from 1.9.1 release 1 to 1.10.1 release 2, I encountered failures detecting its keyboard and mouse on AmigaHID-Pico, an RP2040-based USB-to-Amiga input adapter using TinyUSB.
The Comet detected a full-speed USB connection, but enumeration did not finish. Its USB controller remained at address zero, the Pico showed no devices, and the HID workers repeatedly logged busy/unplugged (write select). Power-cycling the Pico did not reliably recover it.
The logs point to /etc/udev/rules.d/99-otg-reconnect.rules, which runs /usr/sbin/reset_uvc_udc --once when USB connection state changes. This performs a full DWC3 controller unbind/rebind, even with USB webcam emulation disabled and only HID functions active.
In one captured sequence, a reconnect triggered that reset and failed to finish enumeration. Another reconnect six seconds later skipped the reset because of the script's ten-second cooldown; enumeration then completed and both keyboard and mouse worked.
I worked around it by disabling that reconnect rule over SSH:
mv /etc/udev/rules.d/99-otg-reconnect.rules /etc/udev/rules.d/99-otg-reconnect.rules.disabled
udevadm control --reload-rules
Afterward, I rebooted the Comet and AmigaHID-Pico automatically detected its keyboard and mouse. The boot log showed one USB reset followed by USB_STATE=CONFIGURED.
This suggests the extra controller reset can interfere with enumeration on this host. Could it be avoided for normal HID-only reconnects, or performed only after a failed enumeration attempt?
This is a workaround confirmed by a few hours of testing. Disabling the rule also removes its automatic USB recovery behavior. To undo it, rename the file back to .rules and reload the rules.