I created a Scale Set within Azure from a specialized Windows 10 image with some custom software stack installed.
When I try to run pipeline jobs on this, all seem to work fine until suddenly the agent gets killed as unhealthy by Azure Devops. While I tried several things out, I observed, that this arbitrary kills are somehow correlated to running a second job on the scale set. My guess is, that Azure Devops can't handle multiple agents with the same computer name, which is - by default - the case when I create a ScaleSet from a specialized Azure VM image.
Is there any workaround or does Azure Devops only supports ScaleSets from generalized VMs? I actually tried to change the hostname with a custom script extension to the VMSS, but I didn't get it to work.