Increased R1 to lower the cutoff frequency of the DC offset filter (found it unnecessarily distorted the input). I would actually just leave it out altogether.
Changed IC1.2 to non-inverting, removing the need for large impedances in the feedback loop. The inverted and non-inverted outputs are now swapped!
Added a clipping network to IC1.3 (only works in conjunction with the above change!). The output of the module is very hot, almost +/- 12V. The added D2, R12, R13 clip the outputs to a little under +/- 10V. I don't know if there's a consensus about envelope output voltages in Eurorack though (Moog uses 7.5Vpp, Erica synths 10Vpp).
Wired the Release pot in parallel with C8 instead of in series. This gives it more of an actual attack/release behavior. The Attack pot (still) only works semi-properly when the release is at max because the two pots act as a potential divider together. You need to compensate the loss using the gain pot. This design would benefit from an actual slew-rate limiter instead of a peak detector. Check Kassutronics' blog for a nice schema.
Changed RV3 and R5 to 10K and R4 to 100R. This eliminated most of the noise/self-oscillation problems I found and gets rid of the remaining large impedances in feedback loops.
Flipped the LED because, well, the first output is now inverted and the second one is non-inverted 🙂
From https://discord.com/channels/932182348559175730/933072586043170886/954042505811337268