Microbursts are nanoseconds worth of traffic that are so small most monitoring solutions will not see them (they don’t poll often enough).
Instead what you may notice is output queue tail drop.
This becomes important when discussing how interfaces actually transmit data, on a 1 Gbps interface, a 100 Mbps flow is actually the interface transmitting FULL speed (1Gbps) for less than 10th of a second then stopping.
Normally switches have buffers to absorb these, or protocols have built in mechanisms for dealing with them. The issue with these buffers is they can only buffer about 10ms worth of traffic, then they are overloaded. Some switch vendors tot big buffers, but then we get into the issue of holding packets for too long in the buffer and they cause retransmit (buffer bloat).
The reality is there will always be microbursts and there will always be output drops caused by microbursts.
The issue becomes more of a priority when talking about UDP based traffic and no QoS in place.
TCP will retransmit and scale back its window size if it needs to, UDP does not have this.
Imagine the scenario of multiple devices microbursting at the same time, their traffic will make it to the switch just fine because they are on dedicated copper.
The issues start when they are leaving towards to destination, usually this destination interface or interfaces are oversubscribed a certain ration.
Those two microbursts that just came in now want to leave at the exact same time and they fill up the output queue, and tail drop.
These scenarios are why QoS is important, especially with respect to UDP and micro bursts.