The TI parts were made using diffusion. Supposedly more recent domestic FETs have used ion implanation, creating defects in the crystal lattice that are thereafter partially healed by annealing (cooking). According to Ed Oxner, formerly with Siliconix, things have never been the same since.
Toshiba parts like the excellent 2SK170 are still made with diffusion I believe. Their starting material may be purer as well. For really low-capacitance applications it would be interesting if they could made a 4416-like part.
Another feature of the TI parts of significance for super low current/charge amplification: the construction had the FET chip sitting on the top of the gate lead, which comes straight up through the borosilicate sealing glass with a minimum volume of surrounding glass. Other vendors from that era sort of stoomped over the lead and embedded it in the glass, hence with a lot more glass around it. This raised the noise from the lossy glass dielectric. I suspect that losses in a mic capsule make this noise source insignificant.
The different FET part numbers for the same process are just gradeouts I believe.
Process 52 is a higher voltage lower transconductance process. I don't see where it would offer an advantage in this app.
Process 94 (such as the NDF9406) is a dual part. Very low leakage and moderately high voltage, but not particularly low voltage noise.
As mentioned in another thread, there will be a larger contribution to noise from a 1Gohm resistor until the leakage current of the FET gets into the tens of pA range. ~50pA Ig if it is straight shot noise; less depending on where a 1/f spectrum starts to show as you go down in frequency.
