This is interesting. Intralayer caps, but using ceramic, not FR4.
Up to 3 nF/cm^2
They show generic graphs of noise, but nothing of impedances vs frequency. It would be interesting to see just how well it works on a small board.
I saw a test board once that demonstrated that when the chip and the caps are both connected to internal power/ground planes it doesn't matter much where even the high frequency caps are. Moving them away from the chip may add inductance, but as part of a transmission line which supplies current until the pulse reaches the cap.