I wasn't referring to the RIAA EQ, but rather the final system frequency response, as the high end can extend way past 20k. I just didn't know it began to roll off at 15k
What begins to roll-off at about 15k is peak amplitude. The limitation is similar to slew-rate limitation. The higher the frequency, the higher the cutter acceleration, and the higher the power to produce this acceleration Recording a 15kHz sinewave at nominal amplitude would require hundreds of watts in the cutter coils, which it would sustain for not a very long time.
But indeed, the low level response can exceed 20kHz.
When vinyl was just about the only significant media, SE's would know what they sould not do if they wanted a good pressing. The mastering engineer was there to put a final touch, in view of two main factors:
one was the indications of the lathe, that would guide how to apply elliptic EQ (LF monoing) and HF limitation
the other was a touch of artistry, since the monitors in the cutting room were often better than those used for tracking; there were two reasons for that: first, the acoustics of the cutting room were not hampered by the presence of a big desk, second the monitors in the studio were optimized for loudness, in order to impress customers, those in the cutting room were for accuracy.
Today, mixes are produced for digital release; not many SE's know what to do for vinyl cutting, so the vinyl ME has a tougher job to do. I believe Gold will concur.
PS: in the same era, cassette was very much alive. Preparing masters for cassette duplication was a similar job, with also HF limitations and acute noise concerns. It was not associated with the risk of frying a $3000 cutter, though.
Mastering was a technical job, above all, that implied knowing the workings and limitations of the equipment. Today, what they call mastering is a completely different game, with a lot of hype and subjectivity.