Normally, packets are sent "as is", which can contain very large bursts of data (i-frames and p-frames particularly), which can be big enough inturrupt the QoS for other data on the systems with certain network devices, causing them to have to wait to send out, and thus create increased ping. This usually only happens with network devices that aren't really all that great. Higher quality network devices can typically manage the packets better to accommodate for that situation.
What this feature does is segments the packets to certain sizes and sends them out in more spaced intervals to give breathing room for other outgoing data on the system, so those other packets won't get delayed as easily. However, segmenting the data means that it increases the number of TCP acknowledgements that occur, so can increase the chance of frame drops on weaker or less stable connections, hence why it's an option and not always on by default.
