Data Quality and Realism
INSANE separates two kinds of imperfection, and an estimator must treat them differently.
Known Defects
Known Defects are genuine failures — dead sensors, missing scaling, frozen placeholders, swapped columns, maintainer slips. An estimator should detect and exclude or correct them; they should not be learned from.
Desirable Ugliness
Desirable Ugliness is the realistic in-flight difficulty a working sensor produces — asynchronous rates, dropouts, GNSS fix transitions, range spikes, VIO drift, vibration, magnetic disturbance. An estimator discovers and handles it online from the measurements and their reported quality, never from a hardcoded list of where the bad data is.