That's not a bug - you put some vector shapes inside a raster file, and scaled that raster result.
(putting skateboards inside a semi does not mean that the semi can do skateboard tricks)
A vector smart object would need EPS, PDF, or AI data - which would then be rasterized at the specified size/resolution/transformation.
Anything inside a raster file is a raster image, and gets transformed as a raster image.
In short, the "bug" that you are describing is perfectly expected behavior.
I've done this a million times in CS6 and it works perfectly. Vectors should stay vectors. It IS a bug.
No, it hasn't changed at all since I wrote the Smart Object feature.
No matter how many vectors you draw inside a bitmapped image, the result is a bitmap.
Not a bug, just a misunderstanding on your part.
Read this thread -link http://forums.adobe.com/message/3498406#3498406
you may want to read it from the beginning.
Basically while a smart object layer embedded object may contain vector graphics. Photoshop renders a composite for the embedded object and uses the resulting rendered pixels for the smart object layers pixels. When you transform a smart object layer your transforming those pixels like raster layer using your Photoshop preference interpolation preference.. If you want to resize the vector graphics contained in the embedded object you must open the embedded object into the approbate application and resize the embedded object using vector tools. After you commit the resize. Photoshop will update the embedded object and the layers pixel will be the composite rendered for the updated resized embedded object.