The approach that I take, personally, is that filters are the cure for a bad signal, but not for a bad transducer. I will use equalization to fix the recording and not to fix the playback system. If I can process something offline, destructively, in a digital editor, then an FIR filter is great because I don't care about latency. If I want to listen to a record or tape and I need to fix the tone, then an analog EQ is a whole lot more practical than transferring it to digital and fixing it there.
For any source with a microphone involved, there are already phase shifts coming from that first transducer in the recording chain, the mic itself. If the mix engineer turned one EQ knob on the desk, there's more. For a record, the RIAA stage in cutting added some, then the lathe's cutting head, then the cartridge, then the RIAA stage in my system, and then the speakers. None of these are perfectly flat, and they're all introducing some group delay.
It's not fair to say that every record is an example of "phase soup", but it's certainly fair to say that any recording that uses a microphone is provably affected by at least some small amount of group delay in comparison to the original sound in the room. If I can grab and turn a knob that adjusts an analog filter, then, sure, I'm creating more group delay at that point in the chain. But as I get closer to what it sounded like in the room (which is unknowable for me), I'm also undoing some group delay from previous parts of the chain. So, for me, it's not worth worrying about.