With VST-XML (or VSTXML) when you drop the audio event from the arrangement, you actually drop a code snippet on the VST instrument. The code informs the VST plugins where to find the sample file(s) and what regions to play. Also how the file should be treated, like: Tempo, quantize data, root note, region data (slice) etc.
A common single XML slice looks like this:
One of the benefits is you can slice your samples in your host (I use Studio One, but Cubase supports it too) and maintain the slices in the sampler, but still have the possibility to adjust the regions of the sample in the sample player. Also you can simply edit a sliced sample as a single file, because the sample player reads the entire file and maps it according to the regional data. So the file isn’t physically sliced.
It’s pretty convenient.