Hi,
Banks are implemented as 24 arrays of 100 channel numbers. Channel numbers are used as indexes into the channel memories arrays. From the output it appears that one or more of the channel numbers is outside of the bounds 000 - 900.
Since channels 900-999 are actually skip channels, 999 is probably the EAI channel and 1000-1100 are the band scan channels it is conceivable that one or more banks contain legit channels outside of the expected range. Have you added skip channels to a bank, for example?
The patches I've submitted do not seem to touch the code that has the problem here. I've checked with a fresh copy from Hg right now and it works for my images. If you want to share an image that is triggering this I'm happy to characterise the problem more exactly.
I've also been meaning to look into why generating the bank page takes 48 seconds on my CPU. Seems like it's doing something a bit poorly.