We recently experienced the same issue.
The spark job was running fine before and no changes have been made to it before the error suddenly occurred. While we were unable to find the cause of the issue, what did solve it for us was to use a different spark pool. The sole difference being their respective spark version (old: 3.2, new: 3.4).