Hello Bhante @sujato and all.
It seems that the inclusion of the CPED along with NPED had two benefits. First, it helped readers of the EBT with forms that occur as headwords in the CPED, but not in either of the NPED or the PED. Second, it improved the automated lookup of such words (which are now sometimes fails, when with the inclusion of the CPED it succeeded). The main drawback is that such words (as in the cause of akataññū above) the CPED uses alternate spellings of some words
Both both benefits of inclusion of the CPED can be illustrated by the aorist form ahosi, “he, she, or it was”. Without the CPED the aorist form is no longer found. In its place is found a + hosi, “you (sing.) are not”. I suppose that the latter could occur, although I’m not aware that it does. If so, the automated lookup should, in theory, find both.
A native Pali speaker would have discovered and resolved such local ambiguities subconsciously in a context-dependent way that cannot, I think, be easily modelled by lookup alone. In any event, optimal Pali lookup seems a long way off, especially considering the problems posed by sandhi.
In the short term, however, it might be worth considering adding a new lookup mechanism to replace what the CPED provided for word forms where the current lookup mechanism now fails or is incomplete. In many cases, the lookup failed before the CPED was removed. I suggest that the best way to do this is to have the additions of new word forms user-driven.