Observed in Kana's use case: when a prediction validation request is not validated (returning a 422), we do not log what the issue was.
We should log the details of the error (either here or in AtmosphereX) to be able to inform users when such error occurs, but it should not include the values from the request, only the keys which created the error (which I think is the default behaviour from the detail message created by the pydantic validation module).