"sawubona"
I found out that isizulu.net got that from Poulos, who says this:
"As far as concords consisting of consonant plus vowel are concerned the SCs and OCs are absolutely identical. Accordingly, a few differences are found only between the SCs consisting of a vowel only and their counterparts in the OC column.
The first difference is in -ku- a 2nd person singular OC which differs from its SC counterpart which is u-. The appearance of k- cannot be morphonologically explained since what would normally be expected is a glide -w- as is the case with class 3 and 6 (i.e. -wu- and -wa- respectively). The reason for the glide is to prevent vowel juxtaposition between the OC and whatever vocalic element may precede it. In fact there is one instance where -w- actually appears in the 2nd person singular OC. Its frequency is limited to the greeting formula sawubona “hallo” (or literally we see you). The question is why is -wu- then superseded by -ku-? Since this phenomenon is a historical development, it cannot be answered in an analysis which is synchronical in its approach."
Poulos only notes that "sawubona" is the only case of -wu- as the 2SG object marker, and he says that he will not consider the historical development. Since Proto-Bantu's 2SG object marker is reconstructed as *kʊ̀-, and there are no other cases of 2SG -wu- in Zulu, I think it is clear that this -wu- is a reduction of original -ku-.