To reply to your comments, the only thing likely to remain stable in the medium term is the interfaces to low-level libraries. TO expand, while the lower level libraries might be retained in a rewrite of higher-level code, they too are subject to the same pressures. As expanded functionality is required they, too, become more complex up to the point where a rewrite or replacement is needed. Sometimes a replacemement is desirable for other reasons, too - because something better at an even lower level has become available.
(As an aside one of the supposed benefits of Object Orientated languages is that the underlying implementation can be changed while retaining the same interface.)
On the other point I think that you are making a serious mistake in carrying the analogy so far to talk of "compilers". (And arguably misreading the analogy even there - DNA is arguably more akin to an interpreted programming language). I was making a general point on the nature of a design methodology that relies on continually modifying existing designs. An intelligent designer would have the option of "going back to the drawing board" and starting afresh - and would likely find it desirable to do so on occasion. Evolution does not have that option. Therefore if, as it seems, that option has not been taken the evidence favours evolution over intelligent design.