Fixed, it seems!
I tried double checking that the tftp was working fine by installing a tftp client on my PC and trying to pull down the file that the PXE ROM is trying to access before it exists (without any error). This is the initial config file that points to the linuz ramdisk image:
LOCALBOOT 0 KERNEL 56/vmlinuz
APPEND initrd=56/initrd.img ramdisk=10240 rw root=/dev/nfs boot=nfs nfsroot=192.168.80.1:/usr/pluto/diskless/56
I noticed that on the third line "KERNEL" was actually "#ERNEL" - I have certainly never edited this file, and it seemed strange is it looked clear it was really supposed to be KERNEL. So I edited it, and bingo, now it all works again! Guess it must have gotten corrupted... but just one character...
Anyway, as I have noted before in various posts, the actual DHCP component of the boot seems to vary between getting an IP almost instantly, and it taking at least 60 seconds before the PXE boot can continue... In the past, when battling the driver issue, whenever it booted fast, I seemed to end up with driver problems later in the boot sequence. When it took 60 seconds seemed to be a good omen
Ever since this most recent problem occurred, it has been getting the IP instantly, but then bombing out of the PXE ROM and looking for a HDD. As soon as I changed the # to K and rebooted, it takes 60 seconds again, but works. The thing is, at the point it is getting the IP address, it is still in the PXE with no software downloaded, no part of LMCE even touched, much less this config file above, so I don't see how it can influence how quickly it gets an IP address. My networking tells me it can't.... oh well, one more mystery