What a friend has is a 20-Watt driver transformer, with impedances (40K CT to 2.5K CT) better suited for audio in class AB1, maybe AB2.
It would be just like a pro-audio tube power amp, just bigger.
Imagine a slightly smaller but similar transformer coupled circuit from, say a 6SN7 to a quad of 6L6’s in PPP. Set the finals bias point for minimum distortion. Observe maybe just a scooch of grids running positive and plate current barely bouncing under wide-open conditions into a load & scope. See it makes about 60 or 70 Watts before it goes into really raspy distortion.
Now set the bias point and try for class ‘B’ operation. Final grids positive a lot, pulling pretty good current, plate current rising in step with increased audio. Heavy distortion sets in, and you determine the driver stage is running out of steam before hitting the final grids with the necessary poop.
To get more drive, you up the ante with bigger drivers (maybe those 2A3’s or 6L6’s) and a coupling transformer that will spot the maximum power transfer where you need it, in those final grids. Weather they pull current or not, your driver cares not ’cause you designed it to work into sloppy loads via low-impedance tubes that can swing into whatever load it sees.
The triodes will handle it OK by themselves, the pentodes need negative feedback to tackle their inherent high plate impedance and get the ‘regulation or damping’ back.
That redesigned amp now makes about 120 Watts with not too bad total distortion, but it costs you more $ to get the parts for that increased power.
That choice is usually made by just what tubes you have vs. how complex a circuit you need.
Hi-Mu tubes offer a simpler setup, as most of them will work with no safety bias and are thrifty in the drive department, but the commensurate plate load impedance is always high and sometimes tough to get a transformer to match.
Low-Mu tubes are hungry for lots of grid swing voltage, and need that to perform well. On the flip side, they have lower plate impedances, can swing lots more current in the I x E equation, and drive variable loads better.
Big tubes do the same thing little ones do, just bigger.