> examples of low-gain op amps out there?
Nonsense. An Operational Amplifier is an abstract concept and has infinite gain.
With an ideal (or good-enuff) Operational Amplifier, your external components completely determine the response, within the limits of system accuracy. Use 10K and 100K, the gain will BE 10. Use 100K and 1.0uFd, it will integrate with 0.1 second time constant. An integrator must have infinite gain at DC, so you need an infinite gain op-amp. (Or not, if your computational run is short, or the output will be 20Hz high-passed.)
OK, they sell "op amps" with gain that is less than infinity, but may be assumed infinite "for all practical purposes", an assumption you must check in every situation.
Use "perfect" 100K and 100K on an inverter. If the OLG is infinite the gain will be exactly -1. If the OLG is 100, the gain is about -0.99; if it is 10, the gain is about 0.9. As you get to OLG of 3 or 2, the actual gain may be enough off to hear.
The specific point of commercial op-amps is that you can't make a perfect amplifier with exact gain, but you can pick two commercial resistors (or other network) with negligible errors, use them with an amplifier with a LOT of gain, and get low error even when the amp warms/cools or is replaced.
You say your amp has 35dB gain, but when you build a bunch they may come out 33dB, 36dB, depending on the parts. And when you go from winter to summer each one may drift a few more dB. Stack several in series, and the gain setting which worked in December is way-off for the July show. Put that 33dB-37dB amp inside a pair of resistors which give 20dB gain, and the change of gain will be lower. Work it at unity gain, and the gain errors will be around 1%, which even in a 10-amp cascade is hardly nothing to audio.
Obviously you can only get so far with a cascade of unity gain stages. Alright, we'll work at gain of 2 per stage, it will not be 7% in error, typically less than 0.5dB per stage. 5.5dB to 6.0dB. If the complete board needs gain of 30, we need 5 stages. At 0.5dB error per stage, the output could be 2.5dB in error. If you run stereo, Murphy tells us that all the 38dB amps end up in one side, all the 33dB amps in the other side, for a couple-dB error between channels and a strong image-shift. Yes, you can pull modules or trim trimmers. But NYDave in his 1940s radio studio with 4 consoles and dozens of transmitter and monitor paths would go nuts as tubes fail and must be replaced NOW without trim-time. (The real origin of NFB is Bell Telephone systems, thousands of amplifiers in 3,000-mile cascades, routed randomly as available, and any large gain error will echo back down the lines.)
As a summing amp, when designed for unity gain from one input, it will be nearly useless past 30 inputs. There will be negligible effective feedback. You may as well sum passively. True, at 30 inputs there isn't a lot wrong with passive summing (very little interaction unless you disconnect a lot of inputs).
A lot of audio systems have been built with nothing more than 35dB gain-lumps. Fashion changed in the 1970s, but the old ways still work. Depends how clean, consistent, and stable your 35dB gain is, and how fussy you are about computed results versus actual results.
As an active-filter: it should do the low-Q things at unity or low passband gain. A hi-Q filter will have to be trimmed for the specific amp, and some filters just won't gel with 35dB gain.
