For those of you who, in the future, skip to the end of the RESOLVED topics.
Big thanks to Quench, Barakandl, and Inkochnito for being constant readers and providing direction; as well as others who contributed.
This is applicable to Paragon, Dolly Parton, and Harlem Globetrotters that use the same soundcard and U3 prom, but also to other Bally pins that use the same soundcard with different proms.
Two IC's were defective:
U2 Latch
U5 DownCounter
Symptoms:
Sound card is working, but tone sequences sound partially incorrect depending on which notes get triggered.
Drain tone sequence in particular did not exhibit a progressively increasing series of notes (for Paragon)
Also tones were a little unstable and not repeatable, and affected by the logic probe when checking signals around U1/U2/U3
Debugging:
Check U2 latch outputs on the data path between the connector, and U1 and U3 with a logic probe, while playing the test tone sequence. U2 outputs should toggle solidly Hi/Lo with bright alternating colors. On the bad pin, the hi was a solid color, but the lows only flickered the lo indicator, and didn't reflect a solid bright low like other good signals on the chip. Also, adding the probe in the path sometimes affected the tone sequence played (either helped or hindered the low signals becoming more or less "lo"). The failed output (Q3bar pin 15) didn't toggle solidly, while it's non-inverted output (Q3 pin 1) did, as did the other outputs. Replacing the chip stabilized the tones played while in test mode, but didn't correct the actual notes played.
Check U4/U5 outputs with a logic probe that has an additional "pulsing" indicator to indicate that a repeatable signal (like a clock) is being seen. In play mode (not test sequence) triggering a sound (bumper or rollover) will leave the "note" playing on U4/U5/U11 even though you cannot hear it. U4 and U5 pins 1,7,9 and 15 should ALL be pulsing. (Quench actually told me this in an early post but I missed it and didn't check until much later). The note cannot be heard anymore because U7 provides a decaying pulse into the analog circuit that drives the note into the speaker. However after the note has been played, the downcounters keep repeating, so their outputs will look like periodic pulses to the logic probe. U4 pin 6 is the ClockIn which will be a fast pulse. U4 pin 12 sends pulses (slower than clock) into U11 when the downcounters complete a cycle, but will sound the same for any note played, even if it's incorrect. U4 and U5 pins 1,7,9 and 15 (Q1-Q4) are not connected, but are intermediate latches in the downcounters and should also be pulsing like a clock if the counters are working correctly. In my case, all of those pins on U4 were pulsing, but they were not pulsing on U5, some being either stuck hi or lo. This causes U4 to count either too fast or too slow. The final output still pulses and resets the counting sequence, however it has counted incorrectly so the overall note played is incorrect.
Other lessons learned:
Incorrectly replaced U3. Initial shot in the dark since it provides the tones to the rest of the circuit, and the sticker on it was missing so I didn't know which prom it was. However in hindsight the board itself had factory stickers indicating it was the correct version of the board, and all of the B/C jumpers and specific missing resistors and caps for that board variant were correct and didn't looked manually reworked; so no real reason to think the prom was incorrect.
Incorrectly changed U1, thinking that the input (from U2) was toggling correctly, but the U1 output was stuck low. Actually, the input was not toggling sufficiently low to trigger the inverter and the real problem was U2.
As noted above, in most cases the digital circuitry (U1,U2,U3,U4,U5,U11) are still actively playing the last note, even though you don't hear it, which helps debugging. (except if the prior tone sequence ended with the null note)
Data inputs A-D are applied to U3 as provided at the connector because they are inverted twice (ie. Hi on the connector is hi at U3)
Data input E is applied to U3 inverted from the signal at the connector, as it only goes through one inverter (Q3) (ie. Hi on the connector is Lo at U3)
U3 only stores "notes", not whole sequences of sounds. The MPU sends a series of notes that the soundcard plays one at a time. U3 is simply providing a number to count... the higher the number, the lower the frequency of the note played. One "note" is a null signal with a count of only 1 that is sometimes used to generated no tone at the end of a sequence. The counters will keep repeating just the last note played in the sequence, so triggering different sounds (rollover, bumper, etc...) will leave different data on U3 inputs/outputs to help debug... not as good as providing your own specific data, but still aids in debugging the flow of data signals (U1,U2,U3) and the outputs from the prom, and through the downcounters (U4/U5) and divider (U11)
U2 outputs on their own don't have sufficient drive for inputs to U3. The U1 inverting buffers are required to sufficiently drive U3 inputs, and they cannot be "bi-passed" by using the equivalent non-inverted output of U2.
Although U1 and U8 are the same chip, U8 is being used differently in the analog circuit and has no power to it. You cannot re-route signals through spare U8 inverters, in place of a blown U1 inverter.
The MPU shares the data bits A-D between the SoundCard and the Solenoid drivers. By default it is always talking to the SoundCard so U2 is configured to pass through all data to U3, and will only "latch" the last note played whenever the MPU wants to talk to the solenoids (when J4P10 "clock" toggles low).