Unfortunately the current scaling code scales ultimately based on some arbitrary hard coded value. Maintaining consistency across all barcode widths with the current PDF::Reuse API is a bit challenging and would become more so by allowing the user to choose. For example, if one chose a particular size W x H, the bar width tends to become very large as the number of chars in the actual alpha-numeric barcode becomes smaller because PDF::Reuse attempts to maintain W x H. The current Koha code attempts to maintain a reasonable W x H for all possible barcodes. And it does not do even this extremely well. Try generating barcodes for '1' versus '1234567890' to see the problem.

This is all from my memory as its been a while since I've had a look over that code.

Kind Regards,
Chris