> I believe it is the same footprint as 2520 and 990
Bit-o-history: after tube op-amps came Burr-Brown's discrete transistor opamps. B-B defined the module and footprint that many-many people used. B-B usually kept the five/six basic pins (V+, V-, In+, In-, Out, and often Ground) in the same place, but there are exceptions (hardwired followers), and the "other" pins could be most anything the module needed.
> Generally ... a direct replacement
Recall the whole point of opamp design. Instead of making many different amps (hi-gain, buffer, mixer, differencer, integrator/hi-cut), you make one too-super-good amp and add a few passive parts to define what it should do in a specific socket. So at some level "all opamps work the same".
But there's no "Universal Opamp". Tubes need +/-300V and flow 1mA, transistors are far easier at +/15V and can flow 10mA. The early B-B opamps eventually settled to good precision, but schmeared fast signals and made hiss current. Audio-heads demanded good speed with high Pro levels which imply +/-24V and 40mA. The resulting 48V max is a bit past the customary 40V rating of many transistors, so that's a Cost. Someone adds up the cost over 3 yards of desk, and decides to build mostly +/-15Vmax opamps with a few heroic ones at key external interfaces. While you can build a circuit to eat 10V to 60V without change, if you are "told" it will only be used at one specific voltage you may be able to save a buck, which adds up on a big desk. Likewise a resistor long-tail is clean and cheap, works fine for hi-gain amps, summers, and Baxandalls, but won't be happy in a low-gain follower. Also that resistor is sized to the supply: some came with a chart for 12V 15V 18V and 24V use.
And of course a basic audio issue: noise impedance. A transistor can give very low noise, but only over maybe 10:1 of impedances. Low noise around 5K is easy with very early Silicon devices; why the 2N5086 blurb says "low noise". Low noise near 500 would be a better zone for a lot of audio but had to wait for another generation of devices.
The nice thing about the 990 and family is that, in low-Z networks, it is nearly "universal". It eats many different supplies, has gobs of drive, has to have muscle because its noise impedance is low, and the sonics are very excellent, miles ahead of some cruder old modules or most ICs (which didn't have muscle either). If you have 1:2 mike-iron or a lot of 10K mix-resistors, it's real happy. If you have 1:10 iron or a few 100K mix resistors, other op-amps may be a better choice.
And the thing about audio is that if you define THE important-to-you parameters, you can sometimes whomp-up a specific device which nudges-out any "universal" device.
It Depends.