Snapshots are cheaper. We optimize everything we can, of course, but more effects nevertheless means more CPU consumption. Adding a compressor effect to your project means paying for one more effect to process the signal; whereas a snapshot changes the property values of effects that were already present but doesn’t add any additional processing to the signal, making it practically free.
(Technically there is one case in which a snapshot can meaningfully impact CPU consumption, but it isn’t relevant to your use-case.)
Nothing, or so close to nothing that it makes no difference. Scoping additional properties into a snapshot doesn’t measurably increase its CPU consumption.
Of course, adding an equalizer effect to a group bus will increase your project’s resource consumption, but it would do that with or without the snapshot.