Several reasons.

• Schwern has his hands full these days and non-essential stuff is getting (appropriately) put on the back burner. (I say this based on email exchanges with him after I identified a possible off-by-one bug in Sub::Uplevel some weeks ago).
• File::pushd uses a different interface that doesn't rely on manipulating automatically-exported, tied, global variables. (Aren't those anti-patterns? Shouldn't we be discouraging their use?) From that perspective, I'd argue my module brings something new to the table rather than just being a fix to an old module.
• The broken functionality in File::chdir has to do with using @CWD to represent separated components of a directory and changing directories by pushing/popping the array. This breaks on Win32 due to the existence of the volume (e.g. C:\ ). It wasn't just broken -- it was never written to be portable in the first place. None of this functionality was relevant to what I needed, but it meant the module wasn't available via PPM, which I try to stick with for my Win32 perl. It was faster to code up and release a more narrowly targeted module (including a test suite) than to unravel getting the right behavior in File::chdir. (Which I did pull down and examine to see if I could manage a quick fix before I even started.)

I think your standards on your home node about "needless reimplementation" are needlessly high.