Beomaster 8000: Sudden On or Switching to FM Preset Phenomenon - The Story Continues

I am still in head scratching mode regarding the Beomaster 8000 that occasionally turns itself on to a FM preset, usually P1. Despite improving the decoupling of the processors from the power supply by adding 100nF capacitors to their power pins, this Beomaster finally turned itself on to P1 while I was present, and a few days later it switched itself to P1 while it was playing from my iPad on Tape 1. While this was frustrating news, it also meant that I was finally able to reproduce the issue, and that the phenomenon happens independent from where the Beomaster is plugged in (one of my theories in absence of being able to reproduce the phenomenon was that maybe the location of my customer's house is prone to a different level of line voltage fluctuations etc...I was grasping for straws, I know...;-). 

Anyway after it switched itself spontaneously from Tape to P1 I thought that the issue must be originating from the P1 input. A look at the circuit diagram showed that the only way P1 could accidentally trigger itself would be the absence of the 4.7k pull down resistor R1 (as long as no key is pressed on the keypad, the pins are electrically only connected to ground via the pull downs and to the input pins of the uProcessor). The interesting thing about the pull down resistors for the keypad inputs is that they are part of an integrated 7 resistor array with a common pin, i.e. the entire array has only 8 pins. This conveniently allows connecting all 7 resistors to ground with just one solder point. My hypothesis was that the array may have some intermittent issues, permitting some of the keypad inputs to float (which is known to cause erratic input conditions in microprocessors) hence potentially explaining the sudden switching to P1. 

I opened the Beomaster up again and took the uProcessor board out. Before unsoldering the array I removed the slave processor IC3 where it is connected to the input pins for the keypad. I usually do this when I work around the processor ICs since new parts may carry a charge which may accidentally fry some pins on the processor while implanting the part. The black 8 pin flat package in the center of the photo right in front of the IC socket is the array.

I extracted the array. This shows it together with the new part (yellow, Newark 62J2860):

While taking the original array out it occurred to me that the common pin (on the left in the first picture) is soldered into a via. And that made me think that maybe not the array is the problem, but the via, whose intermittency is a common problem on this board. The via is needed since the ground connection of the common pin is made on the composite side of the board, but they only wave soldered the board only from the solder side. This means they had to put vias around such pins that capillary forces would draw the solder into the via and through to the component side of the board, making the connection to the top copper layer. In the 1980s they were apparently not yet able to through-plate vias with electroless copper deposition technology, like it is done today.

I soldered the new array in paying attention to soldering it also on the top side. This shows the final result:

I put the board back in and fired the Beomaster up, and it still worked. So far so good...only in a few weeks we will know if this finally cured the problem.