The pain depends almost entirely on whether you choose low-power indoor or standard-power operation. If the product can live within low-power indoor limits, start there. It avoids the AFC database, geolocation and the extra certification test cases that come with standard power, and for most indoor access points the range penalty is acceptable. Standard power is only worth it when the use case genuinely needs outdoor or higher-EIRP coverage, because the AFC integration, the location-accuracy requirement and the test-lab time then dominate the schedule.
On DFS, plan the channel set before the hardware is frozen. The 6 GHz band has DFS channels in several regions, and a radar detection forces a channel switch plus a channel-availability check that can take a minute or more, during which the client sees a dropout. That is the failure users remember, not the average throughput. In practice, ship with DFS channels disabled by default or clearly marked as optional, keep at least one non-DFS channel as a fallback, and make the switch behaviour visible in the logs so support can tell a radar event from a firmware crash.
Test with a radar simulator early, not only in the certification lab. Feed the receiver the standard radar pulse patterns at the levels in the test plan and confirm that detection, channel move and recovery all happen cleanly, and that false detections from nearby Wi-Fi or non-radar pulses do not trigger unnecessary switches. Also test 5 GHz coexistence on the same board: a 6 GHz front end with weak filtering will desensitise when the 5 GHz radio transmits, and that looks like a DFS problem in the field even though the radio is working correctly.
Finally, budget certification as a separate workstream with its own owner. The RF design can be finished while the AFC interface, the channel database, the test report and the labelling are still incomplete, and those items are what delay shipment. A product that passes the RF tests but has no clean AFC or DFS test evidence is not shippable.