In a conversation recently, someone compared Linux distributions with yeast. When they gather enough momentum and density, they kill themselves. Too much complexity, and the system rapidly becomes unmaintainable. Too many people pulling in too many directions, and it becomes impossible to achieve anything. The more successful projects seem to be those that have a clear leadership and direction. This set me to thinking along similar lines about other projects.
Gentoo - yeast threshold hit. Too many maintainers, too complex. Even I can't understand the network init scripts any more.
Debian - somehow manages to avoid the threshold, due to a very dedicated and singleminded core group. Most of the additions which would normally cause it to reach the yeast threshold happen in forks like Ubuntu.
Red Hat - questionable.
Ubuntu - controlled by a sufficiently directed core of people that the threshold is not hit.
Python - while outsiders may hate the religious fervour which the fanatics seem to exude, this fervour does seem to keep the language on track and directed to the task at hand. On the other hand, it still lacks CPAN.
Java - like some projects listed above, control by a core (commercial) team seems to help. Projects tend to be isolated, and therefore directed and dedicated (except for foundries like codehaus and apache).
Perl 6, Parrot, etc - yeast threshold hit before it even started. Does nearly everything, but totally clobbered by its own features.
Note that projects which have died through natural causes, old age, lack of interest, etc do not count. (TCL, Scheme, Mandrake, ...)