> I've never seen it DC coupled
Depends on the AC ratio you need.
If this were the only NFB in a "stiff" amp, say closed-loop gain of 30, and you DC coupled, you'd put about 300V/30= 10V on the first cathodes, which would upset them.
In this case this NFB loop is probably setting closed-loop gain just a hair less than open-loop gain, say 1000. So 300V DC in puts +0.3V DC on first cathodes....
ah-HAH! Voltages like this are "about right" for 12AX7. It is "fixed bias", but a triode has large tolerance for fix-bias since the plate also influences current. It won't vary nearly as much as a pentode.
Now if we have suitable fix-bias, we don't need "cathode bias resistor". We'd normally run ~~1K here, which cuts 12AX7 gain in half. But with DC already there, we want a much lower resistor. Say 100 ohms. The plate NFB resistor is 50K-100K, acceptable. The tertiary winding (assuming it mirrors output winding) NFB resistor could get as low as 3K, which seems low..... split the difference. Cathode resistors are 300r, from-plate resistors are well over 100K, from-winding resistors are maybe 10K to 60K for gains 1 to 4 (tapering to infinity at gain 7).
'83 plate-cathode voltage will be 0.3V*Mu for the fix-bias, 68K+300r*Mu for the self-bias. du-du-du... assuming 100K plate load for good bandwidth, the plates will sit about halfway up the 300V supply, a good point. We could aim higher for a bit more swing, but the '88 only needs 7V of grid swing and the '83 can do 60V, very ample. We could aim lower for a very slight increase of gain, though hardly worthwhile. If you discover it really sits at 100V or 200V, I'll say "OK", but finding it near 150V is OK.
> E88C grids are lifted 100V above ground with 15K serving as cathode resistors. All decoupling caps can be 100uF...
Generally agree. For a "large system" the high-Gm '88 might like >100uFd for near-nothing loss at 30Hz (rated 1dB but probably better, and we have three other poles).
Whether '88 grid resistors are 270K or 820K matters little, as long as their coupling caps are sized accordingly. They could even be above the "max" rating, because the huge cathode resistor will force design current even with volts of grid leak offset.
Any "sound" is very much about the transformers. It may not be possible to build an amp with all these specs or performance using available modern parts. The best source would be the cellar of old German radio buildings, or whoever salvaged that junk when the stations modernized.