ABROCA Distributions For Algorithmic Bias Assessment: Considerations Around Interpretation