:  first things first - have you repaired/replaced the main relay?

Damn it I rewrote this article and snipped out the bits where I
originally wrote that I resoldered everything on the old main relay's
PCB, saw no change, got a new one, no difference.

(FWIW I *think* the tests rule out the main relay since its second
relay won't ever close if both ECM leads off of it are at +12....)(?)

I've also been all over the grounds like stink on you know
what... Good ground to the body from the main relay and at the two
ground leads on the A connector on the ECM

first things first - have you repaired/replaced the main relay?

Anyone ever repair a funky ECM by opening it up and cleaning up any
bad solder joints?  Will the ECM avoid sending fuel if it sees
something "wrong" at its sensors although there aren't any codes
stored?  Should I toss the ECM at this point?

Long nauseating details and plea follow:

Recap: 94 Accord with the classic main relay (PGM-FI) symptoms (no
fuel pump sound for 2 seconds as required when starting, check engine
light on, can't start) with a twist: It only fails in a cold car when
ambient is around above about 70F/27C, *and* leaving the key in the II
(IGN) position for anywhere from 1 to 4 minutes has always allowed the
second relay and then the fuel pump to kick in, CIL goes off and we're
off.  Never a problem while driving, and if the engine is warm it's
never a problem starting.  It also isn't a problem when the entire car
has cooled to the low 60s/low 20s F/C or below.  Cranking it doesn't
help.

What I've done: I've verified there's no 12v back at the fuel pump
when this happens but for good reasonI've measured battery voltage
at the relay pins 6 (good) and 1 (bad) and back at the ECM A25&B1 and
A7 (these are the connections to the relay) where it's good and bad
(both at Vbatt).  I've jumpered the service connector and the CIL
stays on all the time.  I've got good grounds at the ECM A connector
and the PGM-FI.

What seems to be busted: WTF is up with my ECM?  The skiz show the
fuel pump relay in the PGM-FI won't close unless the ECM pulls down
its A7 (which is PGM-FI pin 1).  It doesn't pull it down, not unless
the car is left in II/IGN for a couple of minutes under the conditions
above.

Some specific questions:

- What else am I missing?  

- Does the ECM look for other things, a sensor in a sane range or
something other than battery voltage coming off of PGM-FI 6, before
it'll turn on the fuel pump for the 2 seconds?

- Is it worth cracking open the ECM to reheat any obviously flakey
solder joints?  Can the ECM be cracked open easily?

- Anyone have a compatible ECM they'd loan me to do a swap check if
I've ruled out everything else?