side note/tale of carrelessness
I accidently put a 560R grid leak resistor in one channel of my amp instead of the traditional 100 or 150kOhm. When I found it I fixed it, and found that the linestage had no problem driving the 560R input impedance of that channel. That's proof enough for me that output z of Guinevere is going to be in the *FEW* hundred Ohm range.

If you want to run one section, I am not sure you can heat only the section in question. We don't need to be wasting heater power at that rate.

Paralleling sections also increases the likleyhood of accidently building an RF oscillator. It also raises the old question about the two sections working together. One 5687 section is plenty IMO.

Option offered, suggest skipping paragraph.

If you want high gain, there is a cousin of the 12B4, the 12A4 which has a mu of 20 ( and is indistinguishable from the outside from the 12B4 ). This option raises an interesting possibility, of getting high and low gain by switching tubes. Op points will be within limits and workable, especially if some tailoring of the OP is done with this in mind.

resume!

I rather use one single section per channel. Extra heater current demands are a big enough issue almost of themselves to nix it. Paralleling questions are another vote against IMO. realizing the limitations of a BBS discussion, you could email me and exchange Telephone numbers and we can chew it at length if you wish.
regards,
Douglas