After replacing the serial terminal (yes it worked a few days before),
I managed to watch the VS3100 M38 boot into NetBSD.

Excerpt from the startup messages:

Nboot
: netbsd

le_softc.sc_r2=0x3a0000
...

NetBSD-1.2D (VS) #8: Sat Apr 12 22:58
###I got it from bertrams place###
...
realmem=16580608
...
backplane0(root)
cpu0 at backplane0: Microvax 3100 (KA41)
###note that the ROM claims its a KA42-B V1.5###
vsbus0 at backplane0
le0 at vsbus0: address 00:00:00:00:08:00
###this is wrong, should be 8:0:2b:31:67:cd, which the mopboot detected ok###
le0: 32 receive buffers, 8 transmit buffers
ncr0 at vsbus0: scsi-id 7
scsibus0 at ncr0: 8 targets
ncr1 at vsbus0: scsi-id 7
scsibus0 at ncr0: 8 targets
sd0 at scsibus1 targ 0 lun 0: [correctly identified Quantum 2 GB here]
root device: le0
nfs_boot: using network intervace 'le0'
revarp failed, error=51
###didn't look up the code, but with 0:0:0:0:8:0, nobody would answer###

At this point, I added a /etc/ethers entry on the rarp server as 0:0:0:0:8:0.

With this, it will boot into single user (need to fetch all of the
snapshot to progress).

Note 1: as stated above, the le hardware address is trashed sometime
after kernel loading, and before the kernel's le0 driver has read it.

Note 2: using the break key or the reset key at the back of the case,
I can't get a >>> anymore if the kernel panics during initialization
(e.g., wrong permission of exported root or swap) after the kernel has
taken over. I need to power cycle.

Hm... I should print the NETTEST output, maybe it helps:

cpu type:	a (CVAX)
board type:	a000004 (vaxstation 3100 model 10-48)
sie data:	4010102
boot dev:	99 (lance)
booted from ESA0
Ethernet: 8:0:2b:31:67:cd 8000

Bertram has copy of Ultrix' ramblings.

Regards,
	Ignatios