Modern digital encoding/decoding generally introduces a level of delay
Um, you mean modern digital encoding/decoding introduces error coding that allows us to use channels that are way too bad for analog comms?
and the potential of packet loss.
Um, you mean if even then the SNR is too bad, it might happen that we lose data, but not as much as for the analog voice communication that wouldn't even work to begin with?
We can put math to this. In fact, we've been putting math to the problem of "how much intellegible information can I transport over a noisy channel" for 80 years now.
The result is that we use channel coding, which we only know halfway decent implementations of for digital transmissions. Hence, making the probability of having a useful communication higher than in analog communications is the whole motivation, and spoiler: it works, and that's why literally any communication system that's not stuck with compatibility does that.
You really want to learn about information theory if you want to understand why – it's a fun subject, and the results are pretty clear! (pun intended.)
It is possible to hear radio static in Apollo communications, but voices remain largely understandable - arguably more understandable than a degraded digital signal.
That's a misconception.
The Apollo voice channel had fantastic SNR by modern standards. From NASA's APOLLO EXPERIENCE REPORT -
VOICE COMMUNICATIONS TECHNIQUES
On the basis of this previous work, the NASA and the prime contractors mutually agreed
on a signal-to-noise standard for the Apollo voice channels of 14 decibels for 90-percent
word intelligibility and 4 decibels for 70-percent word intelligibility when both the sig
nal and the noise powers are expressed in root-mean-square (rms) values.
14 dB is a pretty nice thing, and they were using 3.5 kHz channels. To compare this to a modern bidirectional voice standard that does about the same "target understandability", FreeDV, they'd use 1.2 kHz (so, about one third of the bandwidth) and intelligibility works down to 4 dB SNR – in other words, if we just expand the FreeDV bandwidth to 3.5 kHz (which is pretty trivial, thanks to the modern multicarrier digital structure), we'd get three times the SNR, and the thing would work down basically to 1.75 dB. The "70% intelligibility" at 4 dB is way, way, way worse than FreeDV.
Oh by the way, FreeDV is structured like it is because it also has to deal with multipath propagation – a problem spacecomms will usually not have. Under such idealized conditions, things become waaaaay easier.
I'm sorry I'm a bit verbose on this, but I hear this so often from amateur radio people. I actually work in this field and it is totally beyond me how people that take that hobby seriously can still claim something like that, 40 years after commercial and military reality has proven them wrong.
Could Artemis return to a similar 'analog-era' S-band voice configuration to Apollo, without digital encoding?
Certainly not, because that would be incredibly bad.