I tried that but noticed little impact as it still used the global post process settings from the volume in the level. So instead I have set the post process settings to override the more expensive effects such as lens flares. This had a small but noticeable impact. Further experimentation resulted in now almost reaching 20 fps. (Compared to ~80 without them) For other having similar issues: the main things that helped were reducing dynamic shadows to a minimum and limiting the captures draw distance as far as possible.
More control over the lighting quality of the scene captures, independent from the main scene rendering, would be greatly appreciated in the future.
On a related note: we tried to selectively disable the scene captures for certain players in a multiplayer environment but have not yet been able to successfully do so. Should this be possible? As this would at least solve the problem for players that are not looking at the camera feeds.