I start with the drawings and configurations that exist, and with the people who operate the system every day — they usually know exactly when it goes wrong. Then I measure in the running system: flows and bandwidth, PTP stability, IGMP and multicast behaviour, error counters and buffers on the switches, and what the control layer is actually sending.
You get a report with the findings, the measurements behind them and a proposal for the fix, separating what can be repaired today from what asks for a change in the design. I can carry that out myself, or hand it to your own engineers. Also when the system isn't mine and several suppliers are involved.