2100: selftest failure?

I have a 2100 that I had doing some compiles.  Then power went out.

Knowing that it's usually not losing power that fries electronics but
regaining it at the same time as everything else on that branch circuit
is starting up, producing really dirty power for a second or two, I
turned off the machine room at the breaker box.  Something like half an
hour after power came back, I turned it back on.

Now, the 2100 produces nothing on its serial console, and the LEDs stop
at 11101111, which according to one past message google found for me
has something to do with memory.  I've waited way longer than its
selftest ever took before.

Neither break nor ^C at the serial console do anything.  Pressing the
reset switch (next to the BNC Ethernet) makes the LEDs go all-on; upon
releasing it, they promptly return to 11101111.

Is this normal for any sort of RAM failure (and thus it may be as
simple as reseating the RAM), or does it mean something more
fundamental is wrong, like a toasted RAM controller, or is it
impossible to tell, or what?

Is there any way to get an more informative diagnostics out of it?

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       [email protected]
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B