it looks like you nailed it. In the current setup, I would change an
USB client to close the USB session, and later re-open it again if
rfkill is triggered.
This is a change that has some potential challenges. If the driver simply disconnects the USB when the rfkill occurs, then the Linux driver cannot perform any shut-down or clean-up operations on the device side, leaving the device state corrupted. Even under Linux, "disconnecting" USB devices without clean-up via an emulation layer and attempting to reconnect doesn't always work because the driver assumes a device starts from a known good power-on state, and unfortunately USB doesn't offer any generic way I have seen to force reset on a device. So I think your suggestion would require a manual plug cycle to fix, defeating any advantage.
However, if it is not a pressing problem for you, I would postpone
such a change a bit,
Yes, definitely.
Regards,
Colin