There is MDB client variable group_concat_max_len that controls the max length of GROUP_CONCAT output field. The variable default value differs by distros and can be either 1MB or 3 MB.
The value of this field sets the outcome RG column width here in buildAggregateColumn().
There are multiple places where code makes a new RowGroup and the corresponding RGData using 3MB as its column width and if a user sets the value to a higher value this causes PP to crash.
The suggested solution is to convert GROUP_CONCAT into a StringStore only column and avoid using group_concat_max_len value as a column width.