Unexpected Second-Order Active Low Pass Filter Behavior?

For your interest, here's an alternate circuit that maintains the original capacitor values by connecting C1 to the Av=1 point in the feedback:
Thus the usual equations used to calculate the Sallen-Key filter frequency response for Av=1 can be used, and the response is relatively independent of the output gain, (100Hz stays at the -6db point for the filter values here) as long as the Thevenin parallel impedance resistance of R5 and R6 is keep very low.