Maybe have some direct-debug LEDs/test points, and some more 'complex' full-colour LEDs routed through the spare bits in the display layout?
Regarding using spare bits in the shift registers—this would not work with my preferred direct-chaining approach, as I am assuming that only a single row (and hence a single output across all shift registers) is active at any given time—using one of the unused bits for a status LED would not work.
I could potentially use the 5 unused output groups in my LED drivers however, and leave the common anodes permanently tied to —this is fine, as I would not be multiplexing the status LED. This would only incur extra firmware/routing complexity, but could be considered.
Do I want them to be single-colour, or RGB?
If they are RGB, I can reuse the same BOM component as for my matrix—part consolidation & order quantity advantages
But, if they are RGB, then they will require more data lines (if I want a simple drive method)
I will also need individual current-limiting resistors, and I will only be able to produce the few colour permutations of pure R/G/B (unless I use PWM)
Alternatively, I can use addressable LEDs for my status LEDs?
If I use a single-colour LED, I only require a single GPIO pin