I've spent the last few days cross-compiling, and although the overhead of the build system in question may involve the use of autotools by some projects, my own rant would have something to do with people writing and promoting new tools whilst ignoring use-cases that result in those new tools being completely useless for various groups of people.
I've seen quite a bit of build system evolution and apathy towards cross-compiling in my time - just search the Python bug-tracker for the debris of attempts to get Python's build system to stop doing "I need to run myself on the host now!" and actually work for cross-compilation - and hoping that target devices will be powerful enough for everyone to forget about cross-compilation isn't a useful strategy.
I suppose people can argue that the "anarchy" has not only resulted in the construction of autotools - a dubious claim, since autotools does consider things that later tools ignore or miss, and thus does seem to have had some thought given to its functionality - but also the lack of maintenance of autotools and the proliferation of other tools that don't always measure up. Those things have more to do with the perceived cost of fixing things or developing new things than any particular development strategy, though.