It is known some skins which have not been designed correctly will suffer issues with resizing things.
I find that very offensive Neil. Skins I designed in the past that worked perfectly on all aspects have over the years had several issues with Stardock updates. I have been working for over a year doing updates to fix them! For instance, my Cozy skin RedneckDude posted an image of (the green and orange one), was perfect with Start 10 1.8, no issues. My Zeta skin from a little while ago was perfect too, but using Start 10 1.9, it was screwed up. I take offense to your comment saying skinners are not designing their skins correctly. Maybe Stardock needs some better programmers!
Sorry if that wasn't worded well as I quickly responded without time to investigate further.
I initially thought this was the fairly common issue where skin authors had assumed the right hand side would never change size and thus not set margins correctly on things. It is always important to remember the right hand side does change size depending on OS language. This as you will recall showed up again when we added the "Install updates" text to the shutdown button and it became apparent certain offsets were being used in unexpected ways. There was no proper solution for that and some skins that used offsets to centre the shutdown button may not work 100% as the author originally intended when the shutdown button has longer text.
However I have checked this morning and can reproduce the problem and it is a different issue. Everything is fine at 100% dpi scaling, it is purely broken at higher dpi scalings and this points to a double scaling of certain margins which would be in the code end of WB or Start10 (possibly both).
Oddly it seems to be only showing with some of your skins so I suspect you are setting some margins which are getting multiplied and most skins are leaving the values at 0 thus 0 x anything is 0 and no issues shows up. This explains why the included skins work fine and could explain why we do not have bug reports about this during the long beta process we had for 1.9.
I will look to see this is addressed and ensure our QA team properly test things at alternate DPIs as this should have been caught internally.