@Jupp56 Good work and very interesting!
I agree it seems possible reproduce and to locate the issue more precisely, though there is still a way to go as illustrated by @asdffdsa
Perhaps you can capture the SMB packets with WireShark to find out whether the wrong stats results are due to a bug in the Samba server (Debian) or the SMB client layers (Windows+Go).
You may have to downgrade the communication to SMB-2 to work around the SMB packet encryption in SMB-3. This may make the issue disappear if the issue was indeed introduced in Windows 10 introducing SMB v3.11.
If the issue lies in the SMB client layers, then it may be possible to make a simple proof of concept using a small Windows PowerShell script inspired by the Python script you found. And then use this scripts to reproduce against a frequently used SMB server; preferably a Windows Server.
It is certainly possible to get Microsoft’s attention and a quick response. This forum thread contains a relatively new example. The key ingredients are a simple proof of concept using Microsoft software only and an example showing high probability of data loss for a high number of Microsoft customers (preferably governments or large businesses).