Could be a faulty solder joint or a solder joint with flux or corrosion causing an intermittent short between the signal lines anywhere between the MPU and NVRAM. Isolating is probably easier with throwing a different nvram (or original 5101 RAM with batteries then installed) and seeing if the issue still happens.

The problem you're going to have is there's enough variables (including the 4.7uf cap you added to fix a prior issue with the board booting consistently) that it's going to be difficult to sort out the exact problem without swapping-in known good nvram or RAM and going from there. Nothing that looks suspect from the topside of the NVRAM.. flux looks cleaned up off the board. It's not that it couldn't be possible the NVRAM had an issue, but it's more likely some kind of issue with your MPU board since the 5101 socket was replaced and you were having to add a 4.7uf cap to the board for it to boot consistently.

Just looking at the last picture you posted it looks like the bottom left pad of the 5101 footprint is half missing. Depending how flux was cleaned up top-side and bottom side of the MPU, even flux with a bit of contaminants in it could be causing intermittent issues. Traces are routed between pads under the 5101 socket and are close to the pads.. easy place to create shorts or intermittent failures that make/break under heat if a trace or pad isn't making good contact. Since your game is booting there's not a direct short. The aforementioned reset fix to get the board booting consistently could be playing a part in it -- you could rule that out by trying different nvram or a known working 5101 RAM chip though. In its current form, and with limited tools (multimeter) it's probably going to take quite a bit of time to track down the actual issue and unfortunately not something others can really pin-point easily from afar either.

Either that or see what sending it off to someone that does the repairs would cost. I'm really not sure where price-points are today. Your board looks like a decent starting point.

My comments were more aimed at having seen this kind of thing quite a bit on the forums where due to lack of testing tools and experience someone is burning hours, days and weeks trying to track down an intermittent issue -- and at some point time is money and $50-100 for someone else to repair the board is probably the better bet. Or for the cost of another nvram you can swap it out with a different nvram to see if the issue remains.

I've been without spare boards in the past to swap-in and it's without a doubt the biggest way to waste time trying to track down intermittent issues. When you have spares of various boards, memory, chips that you can swap in it makes it so much easier to narrow things down. Just trying to offer some advice to save you from burning too much time. Been there, done that

So wouldn't over or under voltage on the NVRAM apply to SRAM as well if R11 was out of spec? In other words, upgrading to NVRAM doesn't cause the modification to be any more necessary than when factory RAM was used.. other than there's a chance of voltage going out of spec on a $15 nvram chip versus a $2 RAM chip.

I don't think people without tools or experience need to be too concerned if their R11 isn't looking toasty. Personally I'd view this as -- if your R11 is looking brown or otherwise fatigued, definitely think about replacing it or doing this mod to get rid of any heat or out-of-spec voltages. If you can't do your own board work, at least periodically check it. If it's a machine that will be on 10hrs a day somewhere.. just pay someone to do the mod or replace R11.

Not disagreeing about the usefulness of the mod. I just think some people may get the wrong idea that it's inherent to upgrading to NVRAM because that's when it's being discussed as a recommended mod, when really it's a problem with an aging board and fatigued R11 that will affect *any* early Bally/Stern MPU regardless if it's using RAM or NVRAM.