Obviously, in the real world, the delay is indefinite (users typically are locked onto a single conventional channel or control channel and talkgroup) and maybe (if allowed by the system programmers/admins) scanning other channels/talkgroups with priority given to the selected talkgroup (or maybe another like "dispatch").
Beyond that, it would seem that the factory default delay (hold on talkgroup following a voice call) on a scanner would be the best "generic" setting. After that, the only one that would know is the scanner user who make a change to overcome some type of issue (like missing responses) for the systems they monitor. Of course, making changes - particularly to delay and hold settings - will affect overall performance of scanning multiple systems or sites on a single single system.
Having said that - I tend to leave the delays as is - and if I hear something of interest, I'll manually hold on that talkgroup.