Hmmm. I'm surprised that any serious track racer, either sprint or endurance, would think that any PM that uses accelerometers for cadence (and that includes all pedal-based systems, and PMs like the Stages). Every accelerometer-based PM for which I've analyzed data shows a lag in recording power or cadence or both for standing starts, for low-cadence (which is usually when you have the highest crank torque), and for sudden changes in cadence. You can see that when you analyze the ratio of reported speed to reported cadence. The ratio of reported speed to reported cadence for PMs like the SRM (which uses an old-fashioned reed switch for cadence) when paired with an old-fashioned reed switch speed sensor is much more stable. (I think a corollary is that the new-fangled accelerometer-based speed sensors may also be subject to this, but I haven't seen much data from a bike with a reed switch cadence sensor and an accelerometer speed sensor so I'm not yet sure).
I've worked with a few pursuit racers on their aerodynamic and rolling drag. That's an application where you really do need accuracy across the entire range of speeds and powers that you race at. I've had excellent luck with analyzing data from SRMs; I've had terrible luck with analyzing data from two riders who used Stages. Now when someone asks me for help with measuring their drag, I tell them I can't help them if they used a Stages to collect their data.