The current problem lies in the function TestRowLevelTTLJobRandomEntries.
ColumnTypeIsIndexable will return true for JSON objects making them
part of the test cases. However, in INSERT and SPLIT statements, these
JSON objects have an extra pair of quotes around them which results in
parsing errors. Moreover, altering the flag from FmtBareStrings to
FmtParsable only resulted in the test breaking for other data types.

This is not currently reproducible as using INSERT and SPLITS
on JSON columns, when they are primary indexed, work. The issue is
specific to the manner in which the test case is developed.

Jira issue: CRDB-25878