@Wolferl is right, and his comment made me look at something else. I don’t know how I checked this first, because I had it on my mind: https://www.uradmonitor.com/tools/dashboard-04/?open=11000007

The duty cycle is too high. The firmware reboots the unit when the duty cycle reaches 80% . So this is why it appeared as having network issues, the reboot happened right after Ethernet initialisation (including DHCP lease), just to take everything all over again.
This is not the true duty cycle received by the switching transistor because there is a second blocking transistor that quick closes the gate so it is not dangerous for the parts even if the value seems high.

The model A doesn’t have a voltage multiplier, so the inverter works harder to jump from 5V to 380V needed for the tube. The ferrite core’s permeability is affected by temperature, but I see the weather is chill there. Replacing the inductor would help, but let’s put software first: I can provide a modified firmware that lowers the switching frequency and achieves higher voltage with less effort. It will also remove It approaches the audible spectrum though, as we are already at about 15.5KHz

Please check your mail.

Thanks both for your help and patience!