Discrete op amps will not work in the Emerald for two reasons. First, the S-reg current shunt circuit only supports low currents up to about 30 mA or so, and two discrete op amp modules will usually exceed that by a lot
Second, and most critical, is the input offset currents and/or voltages of these modules are orders of magnitude higher than ICs, where input transistors can be thermally matched and precisely trimmed. The Emerald doesn't have DC blocking caps on the interstage or the inverting inputs, so the DC gain remains high (up until the output coupling cap). If ICs aren't used, the signal will lock up at either the positive or negative rails due to the excessive DC offsets.
I assumed ICs were going to be used from the onset of the design process. Discrete modules are NOT a drop-in replacement.
An issue with the OPA27 is that it has input bias current cancellation circuitry, meaning the current noise will be much worse than the datasheet figure in an asymmetrical circuit (such as any preamp basically!). For MM this is crucial for low noise at higher freqencies as the MM source impedance rises to high levels at higher frequencies (inductance of cartridge), so low current noise is important. NE5534A is still about the best bipolar opamp for an MM preamp, or go to JFET input opamps (negligable current noise) like OPA1656.
I have plots comparing noise spectra for NE5534A and OPA1612 (the dual 1611), using 500mH to simulate a cartridge, in a RIAA preamp:
OPA1612:
And NE5534A:
See how the large hump is very much reduced at higher frequencies, the OPA1612 is noiser from 1kHz upwards, manifesting as higher levels of hiss.