The classic opamp all-pass is in Williams 3,
We're doing that but see some harmonic distortion at high frequency, above 8 MHz in our case. This circuit inverts at low frequencies and has unity gain at high frequency. Opamps have bad CMRR at high frequencies, which is why Jim Williams said "always invert."
There could also be some problems from the finite output impedance of each opamp driving the next stage, which sort of differentiates its output. Our resistances are getting pretty low.
Wouldn't this work?
Circuit B is an inverter at high frequencies, unity gain at low. Gotta try that. We'd just swap some parts on the board and redefine some phase conventions.