> > >And there are also bugs which are hidden by debugging e.g. because the
> > >timing changes.
> > 
> > Do you have numbers? I think it's well understood that there can't be
> > a winning configuration to cover all cases.
> 
> It's just an example based on experiences from working as operating
> system developer for several years. I've only seen problems hidden
> by debugging kernels for other reasons (e.g. different use of
> memory allocators).

Another common one is some functions unintentionally cleaning up the
stack [or other vars] - i.e. the "add a print and it doesn't crash"
phenomenon.