This depends strongly on the priority (A, B, or C) assigned to your project:
Priority A: dynamic scheduling enables priority A projects to observe during the requested observing conditions, as determined at the start of the observation. Scheduling blocks of any length are fine as far as the heuristics in the dynamic scheduler are concerned. However, there may be other considerations in determining the best SB length. At high frequencies the length of time during which the specified phase stability constraint, evaluated at the start of the run, can be expected to be satisfied during the duration of the run needs to be considered. This is a function of the time of day, and time of year. Day time observing may also be limited by commissioning, maintenance, or other activities, so SBs that can end by 8am or start after 5pm (local time) have an advantage.
Priority B: the number of priority B projects approved is designed to fit into the available hours per configuration, but shorter SBs (4 hrs or less) help with scheduling around priority A projects.
Priority C: these are filler projects, and SB duration should be restricted to 0.5 to 1 hr in length (certainly no longer than 2 hrs) in order to have any chance of getting observed