Disaster has struck!
I was investigating anomalies in the growth of generated boards and counted the occurrences of each side in my tile configuration. This allowed me to determine at what scale the growth should be at each level.
Unfortunately, doing so revealed my tile set is corrupt! Two of the sides occur an odd number of times in my data set!
Because this is a tile matching problem, the number of occurrences of each side must be even. This is apparent because each side must match with another side of the same type, requiring two same sides for each edge. Long story short, if any side appears an odd number of times, I can never complete the puzzle.
This corrupts my entire generated board state so far and requires me to reset and with a regenerated tile set. Boo!
I think something else is wrong with the generation algorithm and need to investigate that. For now, I’m just going to go through the process of regenerating my tile set.