I'm debating whether those 100Ω resistors are necessary [R1, R2 & R6] -- and R6 is on the circuit side of C2. Shouldn't it be on the power side of C2? That might actually have something to do with noise being coupled into your audio signal. If, say, IC1 is causing current surges [e.g. when internal gates switch, etc.], that is going to generate large-ish voltages across R6 that could be coupled to IC2 [or vice-versa] -- especially if there are no despiking capacitors [and I don't see any].
The 78Lnn series of regulators do a very good job on their own of rejecting supply noise [worse case 41dB -- which, actually, may not be enough for something like a pre-amp, so maybe the 100Ω resistor is needed on the audio supply leg]. Basically, the combination of the resistor and the capacitors create a Low-Pass filter -- with the goal of passing power with a frequency component as close to DC as possible. It's important to design it to do this for the highest possible amplitude of power variation. In other words, the capacitor(s) must be able to directly supply current, with low enough "ripple" that no appreciable noise can get to the otherside of the regulator. Such a filter is only necessary, if the regulator's own ability to reject noise is inadequate. So, this means, either the supply line is very noisy, or the circuitry being supplied is VERY sensitive [for instance an audio pre-amp]. The fact that relays are involved, makes the former a possibility. --just thinking out loud, here.
I would suggest the elimination of the mechanical relays for some sort of electronic switching solution, BUT that may be difficult, considering analog signals are being switched and must be switched with little or no distortion.
And, an aside: Is D1's role to crowbar any backwards connected power source [i.e. a crude sort of reverse polarity protection] -- which causes less smoke if something like a fuse is involved.? There are better ways. Google "mosfet reverse polarity protection".