I don't think this failure is significant. I have seen failures in the double colon test in other OSs and on scenarios where there are mounted filesystems so timing is an issue or where the locking of files causes slowdown. I think the test itself is fragile and might simply be skipped or ignored here.
Also, I note that GNU make http://http.debian.net/debian/pool/main/m/make-dfsg/make-dfsg_4.1-9.1.diff.gz has some diffs that you might apply here as well since the code bases are largely the same. See also the comments about double colon weirdness logged against GNU make which are probably relevant here.