> the 2510 runs at a much higher rail voltage (+-36V) and the 2520 runs lower (around +-20V). In schematics that I've seen, the 2510 is most often used do unbalance signals from sources and the 2520 is most often used to drive transformers. Is this accurate?
Accurate? Who knows? The guys who designed this stuff are hidden in history.
Reasonable? Probably.
Output: Audio transformers are happiest when low-Z, and in transistors current is cheaper than voltage. So a transformer driver would logically be designed for modest voltage, and the windings adjusted for any load impedance needed. 40V transistors are cheap enough to make this a good target value.
Input: you have to take what the Real World gives you. These days it is rarely over 10VRMS, but back when tubes were fading it was not unknown (though uncommon) to run some lines at +14dBm VU, which is around +30dBm on peaks, which is about 70V peak-peak. Which will just fit inside +/-36V rails. Or to see 10V or 20V of common-mode voltage when things go against you: the common-mode plus the signal could be up in the +/-36V range.