No Zenguan, bouncing individual patches is not recommended. In sequencing, you'll need to listen to the individual tracks and change the sequencing dynamically. If you record one track, then after adding a few tracks, you may decide the first track need changes. Then you'll need to re-record it again. That's very troublesome workflow. Midi is always the way to go. You'll have to tweak the settings the best you can do to get satisfactory sound from all the patches.
Actually, in my workstation days (before moving to software), what I would do is this:
There are usually 2 reverb/effects processors in a workstation (minimum) nowadays. E-piano is usually the only instrument that uses a lot of chorus effect compared to the others. One reverb unit will always be set at chorus (with setting as close as to the original e-piano patch as possible). The other reverb unit will be a normal reverb (eg hall). So, one effects processor is dedicated just for the e-piano, and the reverb is for the entire pallette. Then I would make sure each instrument is tweaked accordingly - eg very minimal (close to zero) reverb for bass, stronger reverb for strings, moderate effect for solo instruments etc.
The other way, if you are going the hardware path, is this. Electro/mechanical keyboard sounds (eg: e-piano, acoustic piano, clavs, organs, wurlis etc) are usually a tax on the polyphony. That's because we usually use 2 hands to play electro/mechanical keyboards sounds. So it would be prudent to use a dedicated hardware sound module just for playing these sounds. The workstation itself can be dedicated for other sequencing instruments. This way, your e-piano/piano etc runs off another external unit which will retain it's patch programming (and you get your "sweet DX7-like E-piano sound" setting), and you'll also have a lower risk of running out of polyphony. That was also what I did in the past.
Or the easier way - swtich everything to software!