What is the Pega recommended way to deal with non-permanent error during agent processing?

Consider situation when agent has to update work or data object which is locked at the processing time.

What is preffered solution to make agent process work later when object may be released?

Standard parameter MaxAttempts for Queue-For-Agent does not work as all attempts may be performed at once.

I was thinking about following approach:

1) Clear message, clear activity status to complete current agent task

2) Queue current task again with parameter MinimumAgeForProcessing > 0