I haven't done extensive testing for rates over 9600 baud, so I can't answer your question. I would actually think that higher speeds would be as bad or maybe worse, since as I recollect Serial Mice usually run at fairly high speeds.
The speeds actually dont really matter for the "legacy mouse detection": Windows sees that there is some activity on the port and when the PNP Enumeration (www.codon.org.uk/~mjg59/pnpcom.rtf) fails, with a high probability Windows assumes that the device might be a serial mouse ;D
Easiest way to prevent this behaviour:
a) Switch to USB* and use a native USB Device Class (eg HID or WinUSB)
b) Switch to USB and use a SerialPort/CDC converter chip.
E.g. ft232r (http://www.ftdichip.com/):
- the .inf can easily be modified to exclude serenum.sys which prevents the mouse detection.
(changing the .inf requires a digital signature for newer Windows version - the driver supplied by ftdi includes serenum.sys by default ;/)
- adapters and adapter-cables (rs232<->usb) with this chips are available from various manufacturers.
- For the hardware manufacturer there is also a very simple solution for upgrading to usb: http://www.ftdichip.com/Support/Documents/AppNotes/AN\_149\_Upgrading%20a%20DB9%20RS232%20Interface%20to%20USB%20Using%20an%20FTDI%20DB9-USB-RS232%20Module.pdf
(I dont work for ftdi, just have used those chips a lot ;D)
c) Modify the scale that it wont output data all the time and only on request.
The software could send something like "STARTMEASURE\r\n" and only after receiving this string the scale outputs some data.
=> If the scale stays quiet during enumeration windows should not be able to detect a mouse.
(*at least as an additional option for "current" computers ;D)