Difficult to pick a snappy title for this so suggestions welcome!
Anyway, I had an issue recently where we needed to load test 100+ machines, all booted from PXE. All machines were fresh out the box and all ethernet adaptors were fully registered in SCCM in the ignore list. In short, there was no reason for any machine not to PXE boot as expected. All went well until the 29th machine and then every machine thereafter refused to PXE boot. However, they did still receive an IP address. Additionally, the message Server response timeout E-18 flashed up very quickly before booting from the SSD.
Digging a little further, the machines that were able to PXE boot fine had to receive an IP address between .2 and .31 to work. Any IP after just refused to work. Having accepted that all was well from a firewall perspective, I was forced to concede that the problem was somehow local.
To cut a long and boring story short, the problem turned out to be a mis-configured subnet mask on the PXE server, in this case 255.255.255.224 instead of the 255.255.255.0 that it was supposed to be. The PXE server itself was allocated an IP within the range of PXE bootable clients and anything outside the range didn’t work. So easy when you know but one expects the basics to be correct so it took a while to track down. Hopefully this will provide some more ideas for anyone else who finds themselves with a similar issues.