Cherrybomb circuit question

[add4]

Hello all,
I was looking at the schematic of the cherrybomb and i have a few questions for those who understand what's going on.
link for reference: http://www.madbeanpedals.com/projects/Cherrybomb/docs/CherryBomb.pdf

Is R25 a feedback loop to tame the first gain stage?
This is how i imagine it works:

emitter and collector of a transistor are out of phase with the base. so if i want a feedback loop for one single gain stage, i can take the output from the collector and feed it back to the base, with some reduction (resistance), frequency selection (cap -> high frequencies are tamed, not the bass) or clipping (diodes, so if the difference between the base and collector voltages is above the conducting limit, the collector output goes back to base, out of phase, effectively canceling it -> 'squareish wave'. This is in the case of a single stage feedback loop as in a gain stage of a big muff for instance

So here are the corollary questions :
- if i want my feedback loop for 2 gain stages like on the cherry bomb, i still want out of phase signal so  i have to take the output of the collector of the second transistor, and feed it, not to the base because then it's in phase, but to the emitter -> out of phase signal? right? Then why not feed it back to the collector and have a 'more direct' cancellation?
- Then what's the purpose of R3? which seems to do exactly the opposite? feeding the emitter of the second transistor to the base of the first one?
What's happening then? i prevent the ... voltages from feeding the first transistor ? like starving it from juicy electrons?
- If i wanted, i could also put caps and diodes i guess?
- Also, how is it different from, say, a feedback look for each stage as in a big muff ? (i guess the same effect could be reached with two different stages, but here it's controlling both at the same time and it's easier to get the result you want?
- It's still taming the first stages, but with a dependency with the second one, right ?

Thanks !