I actually don't have an issue with load times as the system in use is an laptop with an SSD running Win10.
CPU Intel i5 2.6GHz and crappy intel GFX.

My optimised images version doesn't make much impact on the overall distro size, but it seems I can now use the non-potato mode fine but with the shader options off.
It does seem to be worth the effort but is one of those things you won't know for sure til you do it.
I understand it is often engine dependent, and how those resources get used and how many etc.

I should point out I have been optimising resources for games since the 16bit Amiga days, and am usually involved in a few Unreal engine projects.
Optimising is second nature now.

Yeah with the audio always try to mix up into the workspace available, rather than down or the lack of effective bandwidth means audio compromises don't always work.
Tiny spot effects without much variation in sound or pitch for example can be saved at a frequency just above the frequency they actually use. That is the ultimate in audio optimising but is time consuming and not needed these days.
I would still recommend dropping most of the clips to mono which will halve the amount of channels needed for mixing, allowing for the samples to not tread on each other.
The music and complex sounds need the bandwidth of the full DVD 48k stereo, but vocal clips can be at a lower than CD quality without noticeable degradation as the human voice does not carry all those high frequencies. 44k mono would be more than enough, unless the samples have panning and complex ambience.

A charming game indeed. I haven't got very far yet but loving the feel of it.
Possibly because I am running in potato mode I didn't see any fireflies at the end of the first part. I assume I was supposed to see something, but wild swiping grabbed a few

I am having a break from other things and having a go at optimising some of the resources.

PNGGauntlet is yielding very promising results and once I have done all the PNGs it should make a reasonable difference to load times and RAM use as well as distro size.
It uses lossless optimisations so no artefacts creep in. You could get smaller with other tools but I wouldn't recommend it.

I haven't looked at all the audio files, but 96kHz is over the top for game audio files. Perfect for production and mastering, but should be mixed down for the final files. Most peoples audio will be 48kHz or even 44k so we are getting multiple layers of 96K audio mixed down on-demand to the final 2 channel 48k. Extra work not needed by the PC and better done beforehand.
They can be at least half the current quality and save a lot of footprint.
Again this should lead to improved load times and RAM use.

All together it may yield a bit of performance boost, especially for potatoes.