Actually this circuit is fairly easy to understand, as long as you don't get into the more intricate part of circuit design, which might require an EE degree or two. As been pointed out, the 18p and 27p caps are there mostly for phase correction, they don't contribute much to amp's gain in audio band. If the open loop gain is infinity, the closed loop gain is set up by the two feedback resistors, in this case, (27k+1.3k)/1.3k which is roughly 22X. For a full 15 watts power output on 16 ohm tab (where the feedback is taking from), the output voltage is roughly 16V RMS, which translates into the amp's sensitity of 0.7V RMS. Obviously, a tube amp doesn't have infinite open loop gain, which simply means that it just needs a little more input voltage to get the same full output. The SCA-35 has a 82K over 560 ohm feedback network, which translates into roughly 150X closed loop gain. Why almost 10X more gain than ST-35? It is because of the tone control circuit before the gain stage that has about 20 db attenuation when in flat position.
The overall gain of the amp is pretty much determined by these two resistors. The more gain each stage has, the more overall feedback the amp has, hence the lower distortion, until the amp becomes unstable. For a typical tube amp with phase shift and resonance peaks in OPT, fairly small filter/bypass caps and all that, the practical upper limit for global feedback is about 20db. In the ST-35, the first triode has about 50x gain with an unbypassed cathode resistor, the push-pull stage has around 1.5-3x gain to the 16 ohm tab, depending on load impedance match and tube age. That gives roughly 10-15db global feedback. So there is a need for a little more open loop gain to get more feedback to produce better specs.
That leads to the "positive feedback" between the first gain stage and the phase splitter. One can always argue that the extra gain is cancelled by the global feedback, so is the extra distortions induced by the positive feedback, and that the push-pull output stage might have more distortions than the 1st gain stage, so higher gain in the first stage means more feedback for the PP stage, and the net effect is reduction in overall distortion. Obviously,this is just one way of getting some extra gain. You can also get the same effect with partially bypassed cathode resistor as long as you set up the feedback accordingly. Knowing how coporate America works, I tend to think that this is the lowest cost solution that can be implemented on that tiny circuit board, for example, a smaller 0.22uf cap might be one cent or two cheaper (and more reliable in those days) than a relatively large 50uF bypass cap and requires less circuit board real estate (smaller circuit board means smaller chassis and less metal work). Since the 1st triode has a mu of 100, as long as the "positive feedback" is less than 1/100 (in this case 1.3K/150K is less than 100), the circuit will not oscillate.
Since you are going to do something "similar", you obviously don't have to copy exact. Just make sure you know the cause and effect when making changes. As it is, this circuit should have enough gain to drive 12DT5, even without the positive feedback, obviously at the cost of lower overall feedback.
