Caveat- I don’t work for Elektron so these are assumptions based on the piecemeal memories of how OB works at a lower level, my model may be inaccurate.
Elektron (please correct me anyone if this is inaccurate) licenses their USB audio driver for use with their multiple embedded platforms and may have some say over bugfixes for general-purpose computer USB chipsets + their embedded platforms) and to optimize the OB streaming feature improvements / performance, developing OB on top of the stable / predictable driver.
If their changes in driver and OB software are primarily surrounding newer platforms, it is unlikely that an obvious regression would miss testing. Not impossible, but fairly unlikely. As well, if they are not writing the raw USB driver I would imagine that this sees a lot of adoption that would help with many clients potentially reporting concerns with any stable build.
Often there are updates to tool libraries, and between builds of the USB Driver or OB, it is possible that milder performance benefits could occur as well. I also haven’t really checked how often the USB driver gets updated with each release vs other OB components. Perhaps the driver is necessarily updated with with each release, but I haven’t spent any time verifying this.
In the end, if things are working fine as is, it’s unnecessary to chase illusory benefits through upgrading. Do I compulsively do so anyway, beyond a delay for any firmware updates to devices? Yes 
To more directly answer your question-
Some components will be shared, some components will be unique to whatever platform is being fixed. I imagine the common shared components will be stabler than more unique aspects, the new Takt 2/Tone 2 configuration may be a superset of them…
This is one of those things that people need to build their own models around to gauge risk levels they’re interested in, “are they separate or shared architectures” is both yes and no depending on area of the code 
Not 0 risk, but fairly low as far as these things go.
I don’t recall hearing complaints about regressions due solely to newer versions of OB that weren’t also paired with large enough changes to device firmware, though someone might also have different experiences there to offer.
Does that make sense, so far as my understanding goes?