Could be as simple as a bad solder joint in the vicinity of R1 or R3.
I've gone through this design before. It's kind of neither fish nor flesh:
The LM4562 has higher input current noise than you'd really want in an MM preamp, so you're going to see higher than average high-frequency noise with a real MM cartridge, and feedback resistors R3/R4 are chosen higher than they would have to be for this opamp with fairly good output driving capabilities, compromising low-frequency noise.
Input filtering is basic, but at least you can install your own choice of input capacitor instead of R23 to match the cartridge / cable combo for best frequency response flatness.
At the same time, when configured for 60 dB the thing is at least 10 dB noisier than an average competent MC preamp (let alone a good one, at which point you can probably add another 6-10 dB). Neither opamp voltage noise level nor resistor values are terribly ideal for this usage. (Gain being spread evenly across all 3 stages does not help.) It probably would have been a good idea to expend an extra two transistors and a couple passives per channel for an (optional) MC prepre, the way a number of other designs are doing it.
It would be interesting to have a graph of maximum output level vs. frequency for both an amp like this and a regular "active EQ" design.