I am going to very some more information about this and I will be back as soon as possible.
Thank you Kevin.
I should have added in the OP that sysprep is run with the /generalize switch.
After sysprep has completed and the NUC has powered down, deleting the DHCP entry for that particular machine then allows a successful PXE boot.
This isn't a fix, but a workaround. I think Intel still need to investigate this.
Thank you for sharing this information. I will pass it to our development team.
I am sorry for the delay on providing an answer. We are checking what can be causing this issue. I will be back as soon as possible with more updates.
I checked on the network adapter specifications and also ran a quick issue duplication using PXE and found no issues, see Technical Product Specification for further details.
I believe this behavior could be related to Sysprep*. I also did some quick check on Internet and found similar reference on other forums (non Intel related).
See for instance:
Those are not Intel own websites and Intel has no control over any content but their pointing at a similar issue. I would recommend that you check on Sysprep*, also it might be of some value if you have a chance to run the same PXE environment with no Sysprep and see the results, that would help out isolate the issue.
As additional reference, there is a never BIOS you may be interested in upgrading your system. See BIOS Update Version 35