I ran some sims on tube-cad, and found that an SRPP driver stage should be better than paralleled triodes and LED biasing.
The benefits are that the lower RK can be un-bypassed, which will preserve the inherent NFB of the stage, and that the output
driving impedance will be much lower, which is good when driving paralleled 1626s.
Typical SRPP driver circuit, as used in SorenJ's 6AV5GA project, everything to the left side of the 0.47 uF coupling cap:
For 12AV7, at your originally specified VBB+ 262VDC, and 6.36 mA, RA and RK should be 400 ohms, 390 or 430 ohms is good.
Stage voltage gain will be 20.
If you need higher gain, bypass RK with 1-2 uF / 10V (or greater) cap, for operating it with a 2 KHz LF roll-off. 100 uF or greater for full range response.
I also ran SRPP sims on 12AT7, 262V, 2.1 mA, RA & RK will be 680 ohms, stage voltage gain is 28, unbypassed.
For 5965 (one of my favorite SRPP driver tubes) 262V, 2.5 mA, RA & RK = 1K, stage gain will be 22, unbypassed.
My own "standard SRPP" stage, I run at 30 to 325 V B+, and RA & RK =1K, this allows using 12AX7, 12AT7, 12AU7, 5965, 5751, 6CG7, 12BH7,
just about any 9 pin twin triode tube (other than 6DJ8) with the 9A or 9AJ base pinout.
6DJ8 is a good tube, just needs to observe the lower maximum plate voltage of 130VDC, this would be good with a 200VDC supply.
6/13/2013 Update/Addition:
One important thing not shown in the SRPP circuit, is that the filaments should be "biased" (floated) at approximately 1/4 of the stage's B+ supply, in this case it is 310VDC, so approx. 78 V would be correct, this isn't really critical, anything between 75 and 80V is more than "good enough". The most easy way to do this is to make up a 1:4 voltage divider, I'd use a 300K 1W resistor in series with a 100K 1W resistor, connect a small (1 or 2 uF, 200V) cap across the 100K resistor, and take the resultant 78VDC and connect that to the" filament winding CT, or if there is no CT, connect two 100 ohm resistors in series across the filament, and then use the junction between the two resistors as a "virtual center tap. This balances the absolute potential difference between the cathodes and the heater - the delta can be either plus or minus voltage, and it also serves to reduce hum and noise.
The 1626 filaments could also be "floated" to the same voltage, if only a single filament winding is available. noise and other stage instability issues. This filament "biasing" draws virtually zero current, and the voltage divider resistors consume only about a milliamp, or less.
HTH
/ed B