I just ran this test of the effect of increasing the capacitance on the ADJ pin of the LM317 to see what may be optimal. There's little marginal utility in increasing the value above 1uF to 10uF. I could have used more capacitors of incremental values, but that would take more time than is worth investing.
Attachments
With 10uF on the ADJ pin AND the Wenzel "Finesse Voltage Regulator Noise" circuit, the noise drops to 40nV/Rt Hz at 1kHz.