Sidechain multi-band compressing instrument mix keyed on vocal

[not-relevant]

I'm trying to make room for a vocal in a busy mix, using multi-band compression/dynamic-EQ on the instrument mix-buss keyed to the vocal via a sidechain.

I've already carefully EQ'd other tracks to leave space for vocal forments and siblence using a spectrum analyser. On top of that I've also tried simple sidechain compression of the instruments mix-buss keyed on the vocal. However, on this track, depending on the compression settings these techniques are either not effective, are too obvious, or have too much of an effect on the rest of the mix.

The vocal is already multi-band compressed/dynamic-EQ'd using Waves F6, in order to reign in some nasty siblence and improve intelligibility. I'm now trying to use the F6 on the instrument buss as well, with bands tuned to, and S/C'd from, the vocal. I'm still struggling to get a satisfactory result, and I suspect the compression from the two instances of F6 is interacting somehow.

Can any one advise me:
- Should my sidechain send from the vocal track to the instrument buss's F6 come before or after the F6 applied to the vocal?
- Should the bands in the F6 on the instrument mix be keyed to the whole sidechain signal? or to the filtered signal? I understand the difference but, in this case, do I want to duck the vocal frequences in the instruments mix based on vocal signal falling within those bands only, or for vocal signal across the spectrum?
- I'm guessing I want mid/side mode, and to only compress mid?
- Would there be a benefit to treating instruments individually, rather than trying to duck the whole instrument mix at once?
- The instrument mix has a couple of quite heavily processed guitars, keys, and delay + reverb. I'm assuming this broad frequency stuff is the main thing competing with the vocal. Drums + bass are on separate busses, and I'm not attempting to duck them. Should I consider applying simlar processing to the drums + bass as well?

I've got a reasonable understanding of what I'm doing here, so I'm looking for advice specific to using multiband sidechain compression to dynamically open up particular frequency bands in the mix.

Thanks