Weak networks: why video quality can drop but messages cannot
Real-time audio/video and messaging tolerate weak networks very differently. Understanding the difference is what makes a correct degradation strategy.
On a subway or in a lift, the network is neither up nor down. Bandwidth swings, packets drop, latency spikes.
Messaging and media should react in opposite ways.
Messages: correctness first, latency can wait
A message arriving three seconds late is acceptable. A lost or duplicated message makes users distrust the system. So the messaging path retries, shows “sending”, and guarantees eventual consistency.
Media: continuity first, fidelity can wait
Nobody complains about a slightly soft picture. Everybody complains when the call freezes. The media plane therefore lowers bitrate, drops frames in order and protects audio first.
This is also why the media plane is scheduled separately: its degradation policy is the opposite of messaging, and coupling them drags both down.
A test for buyers
Do not rely on a headline latency figure. Ask three questions: how does media degrade on a weak network, how are messages backfilled after reconnection, and are the two policies independent?
Products that answer all three usually behave well in the field.