"Debug normally" isn't the case with BigTable or any other MapReduce implementation, is it? Do we think Google and big bunch of others made a mistake with it? Obviously not.
 This paper seems to support that: http://int3.de/res/GfsMapReduce/GfsMapReducePaper.pdf
That seems like dubious logic.
Also unlike Google's bet on commodity hardware, I don't see the millions of dollars of savings to be had if the micro-kernel idea works out. At best you'll get an incremental performance benefit and easier debugging. It'd be different if there were no open source OSes and so rolling your own was the only alternative to paying millions in licensing fees.