Continuous (sequence or random) containers automatically trigger playback of the next sound when the previous sound stops. Meanwhile, the "Under Volume Threshold" behavior is always applied on individual sounds. Setting a behavior on a container does not mean that this container obeys this rule. It simply means that all their sounds obey that rule, unless they override this setting.
When a sound falls under volume threshold, it is either killed, sent to virtual voice, or left playing, as dictated by its "Under Volume Threshold" property (defined in the Advanced Settings tab of the object explorer).
The "Under Volume Threshold" behavior of a sound never has any impact on the high-level structure that launched that sound. This also applies to the "Kill Voice" behavior. In other words, continuous containers are not stopped when one of their sound falls under threshold, even if their behavior is "Kill Voice". In such a case though, the sound engine could enter an infinite loop, instantiating and killing a new sound all the time. In order to avoid this, the sound engine waits for a magic duration of about 200 ms before trying to start playback of another sound.