So the shot-in-the-dark hit the target. Changing the tiny C3 100uF/10V capacitor (5mm diameter, 8mm height) significantly changed the signal behavior...as in it now works.
The only small cap I had that would fit was a 22uF (instead of 100uF) so the delay range from yellow-to-green is only about 3-8 seconds with trimpot adjustment.
However, it still has a quirky behavior that sometimes it will power up stuck with Red LED on and trigger has no effect. It seems to only do this if power is applied when the signal is triggered (green wire NOT connected to yellow wire). If power is removed and green wire is connected to yellow wire, it will still power up stuck in red. You have to remove power for quite some time - many minutes - before re-applying power for signal to power up with green LED.
Anyway, here's a scope shot of the relevant signal voltage identified in the schematic in previous post:
The fuzz on the signal is the 60 Hz ripple as DC voltage regulation is from a zener diode. In above scope shot, the yellow-to-green delay was about 4 seconds.
Because of the occasional "lock-up" in red when applying power I don't think it's worth pursuing as this must be something going on inside the blob on this particular unit. That is, while the waveform above seems to be working, it could be that this is not correct and just appears to be working but not as-designed.
I suspect there are a bunch of these signals sitting in boxes or serving as non-functional layout ornaments.
Bottom line. If anyone has this signal and it is mis-behaving as originally described, I recommend removing the board and using an external control board or other electronics to directly drive the signal head LEDs. Other guys report doing this. If for whatever reason you absolutely need the electronics to fit inside the signal base, I think it could be done with, say, $5 of low-cost components but would require design of a printed-circuit board using surface mount components to simplify matters. I noticed the Lionel board is thinner than normal (1/32" vs. the more common 1/16" thick) so space is at a premium. I believe Lionel already has some signal boards that might even perform this function but note that to fit in the CPS, the trimpot must be located as shown (unlikely in a different design) and the board is trapezoidal to fit the sloped base. And if doing so I'd suggest "fixing" the triggering to be compatible with insulated-rail triggering so that you don't need a 153C or 153IR. The manual I found for this signal is dated 2004. I think insulated-rail triggering was already in use at that time.