A recurring pattern in RTC post-mortems: teams migrate SFUs chasing quality gains that were actually a visibility problem all along. The piece makes the case for instrumenting getStats(), QoE metrics, and real user monitoring before assuming the transport is the bottleneck.