> it seems like the output of line level devices ranges from 500 ohms to some around 70k ohms. Maybe even higher.
Hmmm. Except the eraly versions of the Dynaco tube hi-fi preamp, and some old DIN standard, and many guitar FX pedals, everything is "low" output impedance. ARP modular synths used a 1K output, most hi-fi and "consumer studio" is 47-470, most broadcast is under 60 unless it is true 600.
Don't confuse output impedance with drive ability. A hi-fi box may be 47 ohms output impedance, but be unable to drive full level into anything less than 2K. The miracle of negative feedback.
Mostly we make inputs 10K or higher because that way we just don't have to care if the source is near-zero or 600 (or even 1K). And because most sources found in a studio chain will drive 10K easily.
> how would I calculate decibel loss and makeup gain?
You could just avoid it. Don't try to do Voltage Divider and dB lesson at the same time.
The passive mixer is a voltage divider with a common "lower" resistor for all channels. In the simplest lowest-loss form, the "lower" resistor for one channel is all the other mix-resistors (and their sources; prove that this can be negligible) in parallel.
So for the made-up case of an 11-input mixer with 10K mix-resistors, with all inputs connected, one input sees a 10K "top" resistor and ten other 10K resistors "in parallel to ground". This makes a 10K/10= 1K "bottom" resistor.
The voltage divider is then 10K and 1K. The output is, at a glance, a little less than 1/10th of the input. It is really 1/11 of the input. And so we see that the "gain" of a mix network with N inputs, expressed as a fraction, is 1/N. To bring the overall gain back to unity (for one input), we need a gain of N. A 64-input mix network needs a make-up amp with a gain of 64.
The output impedance, which the make-up amp input sees, of that 11-input 10K mixer is evidently all 11 of the 10K resistors in parallel. This is 10K/11 or 909 ohms. In general, it is R/N.
This will lead you to a spec for a purpose-made make-up amp.
There is a fashion to use, instead of a special amp, a mike preamp. Such things like to see a source impedance of 100-200 ohms, and usually like to run a gain of around 40dB which is, in fractions, a gain of 100/1. Therefore we really want a mix-network gain of 1/100 and an output impedance of say 150 ohms. Using the 1/N and R/N formulas from above, and working backward, we want a network with 100 inputs of 15K each. Gain is 1/N or 1/100, which when made-up with a 40dB or 100/1 amp comes to unity. Mix network output impedance is R/N or 15K/100 or 150 ohms. The only "problem" is that we may not want 100 inputs. But we can build the 100-input version on paper, assign the 16 inputs we need, take the other 100-16= 84 resistors and compute their parallel resistance, then trade them in on one "swamping" resistor of that value. 15K/84= 178.6 ohms, 1dB errors won't kill us, in simple dividers 10% error is less than 1dB, use 168 or 175 ohms.
Prove to yourself, with simple fractions, that a 16-in 15K mix network with 178 ohms swamping the bus gives gain of 1/100 for any input and an output impedance near 150 ohms.
All that dB do is turn all this gain/loss multiply/divide into add/subtract. When you need to compute the gain/loss of a phone line from here to Podunk, 13 lossy segments with 12 booster amps between, and do not have a calculator, add/subtract dB is much easier than multiply/divide fractions. But for simple work, it is no big deal. And for voltage dividers, fractions are the easy logical way to solve the problem and the dB number is just for the spec-sheet.