A visual stimulus must last longer than some threshold duration to be perceived consciously. For example, a light dot flashed for a duration of 10ms can't be consciously perceived. Why? What prevents a sub-threshold flash to be consciously perceived? Does this mean the relevant neurocircuits have to keep on firing for some threshold time in order to reach consciousness? If so, why?
If by consciously perceived, you mean recognizing a feature (and possible reactions accordingly), there are several tasks to be considered that seem to cause delays:
Saccades and eye movements: During saccades we are blind, so we must wait a little to bypass a saccade so participants are able to see. It is called saccade masking.
Transmission: neurons are not that fast (around 100m/s for myelinated ones), so transmitting action potential from retina to the primary visual cortex takes time, of around hundreds of milliseconds.
From visual cortex to other area of the brain (for verbal responses, reasoning, matching, or further recognitions that are not simply visual features),
Transmitting reaction signals from motor cortex to the hand so participants can respond.
These are just neurological pathways. Considering some other cognitive shortcuts (e.g. schemas and categories), participants need to bypass their default reasoning modes (unconscious?) that are learned by training, stereotypes, or being stimulated so long that it's been moved from foreground to the context (speaking of perception without awareness).
Depending on how you define stimulus duration, the auditory system likely has no minimum stimulus duration needed for conscious perception. We are quite sensitive to "clicks". A "standard" click duration of 20 microseconds (a single sample at 50 kHz) is not uncommon. Decreasing the duration below this would simply require better stimulus generation hardware (faster sample rate and higher bandwidth transducers). Decreasing the duration would simply add high frequency energy that falls outside the audible range.
An ideal impulse with infinite bandwidth will have a duration of zero (and a magnitude that is a pain to describe), but a real transducer with finite bandwidth will result in some finite duration. Further the stimulus will be mechanically filtered by the outer, middle, and inner ear further decreasing the bandwidth (aka increasing the duration). Finally in the inner ear the signal will be transformed from a mechanical/acoustic signal to an electrical signal with a duration that depends on the place in the brain that you measure the electrical signal.