ScreenCamera.DepthTextureMode and PhotoCamera.DepthTextureMode don't enable depth generation on cameras attached to an avatar.
Instead, avatar cameras use a previous [and invalid] depth buffer rendered by ScreenCamera or PhotoCamera, which leads to off-axis depth effects in shaders rendered with these cameras.
This produces artifacts and broken shader behavior with third party camera systems like VRCLens, etc, in worlds that utilize DepthTextureMode instead of a depth light.