The service manual says that Z103, along with R105/6/7 and Q103 are used as a constant current source. I think that a capacitor in parallel with Z103 will do little to improve the circuit, and may degrade it. On the other hand, a capacitor in parallel with R105 should help to remove some zener induced noise from the current source. Since the current in this section of the circuit is quite small, I'd use a capacitor in the range of 1-10uF. Also, I'd use a polypropylene film capacitor.
In my Aleph amp I by-passed the zener diode with a small plastic capacitor, it seems to work well, reducing noise: this value should not to be too big, in order to avoid any transient during the power-on.
As Thoth said, few uF are a perfect choice, in my opinion.
What really mattes is does it improve sound or degrade it. Can you really hear a difference with or without it? Let your ears be the judge. I really doubt that it matters as it would have beem built in in the first place.
Salt Lake City, UT
in my experience, in other contexts, I appreciated zener bypass, both through instruments, both ears, that is more important.
In this case actually I've not yet tried without bypassing this diode, but this should not be an exception, since a noisy Vgs produces a noisy drain current.
The nfb probably deaden this, but maybe it's better to soften it previously, at a very low cost.
However, as soon as possible, I'll try to alternate the two layouts, to make me sure of this.
On this forum, many diy proposed other ways just to avoid zener noise in this position, and they had been around a lot.
Thirty bilions flies eat ****, they can't be wrong? I don't think, really.