Sharing one antenna works, but only if you verify the switch and the arbitration on the bench rather than trusting the datasheet timing diagram. The failure I see most often is not the arbitration logic itself — it is the RF switch and the desense it causes.
First, measure the switch where it actually operates. Put the switch on a VNA and check insertion loss and isolation across both bands with the real control voltages applied, not just at the nominal state. Isolation from the Bluetooth transmit path into the Wi-Fi receive path matters most: if the Bluetooth transmitter leaks even -20 dBm into the Wi-Fi LNA while Wi-Fi is receiving, the Wi-Fi sensitivity collapses and no amount of arbitration tuning will fix it. A good acceptance test is Wi-Fi sensitivity measured with the Bluetooth transmitter running a continuous SCO link, compared against Bluetooth idle. If sensitivity degrades by more than a couple of dB, you have a leakage problem, not an arbitration problem.
Second, check the switch timing itself. Use a scope on the switch control lines alongside a spectrum analyser on the antenna port and confirm the switch actually moves within the blanking window the arbitration assumes. I have seen arbitration tables that assume a 2 microsecond switch time while the hardware took 10, so every handover clipped the start of the Wi-Fi packet. The fix was a longer guard interval in the arbitration table, not a new switch.
Third, check the antenna match in both switch states. The antenna sees a different load depending on which radio is connected, so measure return loss with each radio's matching network switched in. A shared antenna tuned for Wi-Fi can be several dB worse for Bluetooth, and the result looks like poor Bluetooth range when it is really a matching problem.
My practical order: measure switch isolation and loss, then desense with the other radio transmitting, then timing, then the arbitration table. If the first three pass and throughput still drops, only then tune the arbitration.