A couple of comments:
1. Typically, re-administration at less than 24 months apart is not considered to be valid, with the main risk inflated scores. However, it usually affects the VSI and FRI the most, so that might not be a big factor here.
2. Preschool children are notoriously difficult to test accurately, due to their highly variable attention, motivation, and engagement. This is one of the reasons scores are not considered to be terribly stable until around age 9 or so.
3. A couple of the subtests administered change from the age 3 to the age 4 level of the test.
In VCI: At age 3, the index consists of Information and Receptive Vocabulary (a word is given, and the child selects the matching picture from a number of choices). At age 4, it is Information and Similarities, with the latter more of a verbal reasoning task than a verbal knowledge task. Hence, not an apples to apples comparison of VCIs.
In VSI: At age 3, the Block Design subtest uses solid-colored blocks. At age 4, there are also items with half-colored blocks.
4. Children, especially preschool children, also undergo sporadic periods of rapid cognitive development, usually in only one or a few areas at a time, not across the board. If testing captures the moment just before or just after one of these growth spurts, it can skew the results. This is another one of the reasons that preschool testing is not considered stable.
5. Possibly an example of #4: if your son, by any chance, learned to read between administrations of the WPPSI, that could hugely affect his VCI, as his access to vocabulary and language would have expanded in ways not available to the vast majority of preschoolers.
Do you have any information on subtest performance? Test behavior observations? Anything that changed or occurred in life or school?