I agree entirely, which makes question even more why they've kept it in their despite it apparently being reported as a bug so often.

You're right about 'harmless' uninitialised reads obscuring real problems, C++ copy constructors are particularly bad for generating these when the original constructor doesn't initialise everything.