The man himself has put in his two cents worth about the state of current kernel development and he seems to agree with Andrew Morton that there are too many bugs in the current Linux kernel. Of course, one bug is too many, but by any realistic standard, the current kernel has some glaring defects in versions for several architectures. I'm having enough issues here on one x86 and two PPC machines that I have rolled back to an earlier version on the PPC boxes and instituted a couple workarounds and given up on hot-plugging for now on the x86 machine.
The whole thing seems rather laudable to me. Admit the defects and set aside time and effort to fix them. It's interesting that when Oracle, BEA or Microsoft (all of whom face similar issues with upcoming releases) do the same thing (with or without delays) they get hammered in the press.