The smaller you make the audio buffer, the more susceptible it is to interference from other processes running on the computer, and it generally requires a faster system to work well.
DPC Latency starts to become a factor too.
Latency settings in the driver are for when you need low latency for realtime applications (e.g. monitoring audio) or you need to have a known fixed delay.
For example, I know that when outputting 192kHz my DAC takes 0.47ms for processing, and if I use an 8192 sample buffer, that's 42.67ms - so to correct audio/video sync, I need ~43ms correction. Actually, it's quite a bit more complicated than that, but at least I now know the DAC has a fixed latency which I can factor out. (Media Center itself has some latency, and so does the display)
If you are just using the DAC for music playback, you should be using the largest buffers possible, so that it is not susceptible to interference.
There should not be any audible difference when you change buffer size, up to the point that it is too small for audio to be played back correctly. My DAC allows you to set the buffer from 64 samples to 8192, and at lower values it starts sounding distorted when the buffer size is too small, and if you go even lower once you start hearing distortion,
then it starts dropping out.
You will probably also find that higher sample rates require larger buffer sizes for stable audio playback.
The Meitner MA-1 website seems to offer a native ASIO driver provided by Thesycon, which I would recommend using rather than Kernel streaming.