Dear colleagues,
Just to add some fuel in the discussion of understimation of reliability, Applied Psychological Measurement published just an article discussing of the attenuation and deflation in the estimates. While we usually refer to attenuation of 1 - 11%, the true underestimation may be .60 - .70 units of reliability or 40 - 60%. Some options to rectify this are offered in the article.
https://www.researchgate.net/publication/363566401_Attenuation-Corrected_Estimators_of_Reliability