One of the strange things about AV is that we spend a great deal of effort reducing latency, then turn around and add delay on purpose. Both efforts serve the same goal, which is getting everything to arrive at the right time.
Delay speakers are the easiest example. Sound travels at roughly 1.13 feet per millisecond, so if a delay loudspeaker sits 100 feet into the room from the mains, the sound from the main system takes about 88 milliseconds to reach it. The electrical signal feeding the delay speaker gets there almost immediately, so we hold it back, usually setting it slightly behind the arrival of the mains. That small offset keeps the listener hearing the sound from the stage instead of from the loudspeaker hanging nearby.
Video is another example. Audio moves through a system quickly, and video often cannot. Frames have to be buffered and processed, and at 60 frames per second every frame represents 16.7 milliseconds. Stack a few frames of processing into a system and the audio arrives noticeably ahead of the picture, which audiences catch far sooner than audio that lags. The audio path is already faster than it needs to be, so we add delay to the audio.
Networked audio presents a subtler version of the same decision. Most platforms let us choose how aggressively to minimize receive latency. The minimum setting looks attractive on a specification sheet but leaves the least margin for packets that arrive later than expected.
I was involved in a project that carried telemetry, audio, and video across a medical campus for patient monitoring, and we made it slower on purpose. The physician could tolerate a couple of seconds of latency. What was mandatory was that all 3 streams arrived together and stayed together. Telemetry was the fastest, and video was the slowest. We delayed the audio and the telemetry to match the video, then raised the latency settings on every stream so each one had enough buffer to hold its delay constant.
The lowest-latency setting is often the wrong setting.