I am not sure. Does it happen on a continuous basis? Can you ping those targets directly -- with the same configuration (interval, timeout)? If yes, may be there is an obscure bug somewhere in cloudprober (I say obscure because we run a lot of ping probes with a lot of targets, and haven't come across a bug that looks like this -- we see duplicate replies sometimes, but they are explainable).
Also, can you take a tcpdump to see what's going on? Are there really duplicates on the wire? May be cloudprober is sending duplicate requests (requests with same ICMP ID and Sequence number) due to some bug. Cloudprober continues to use same ICMP ID throughout its lifetime and cycles through sequence numbers. So if you're sending 2 packets every 2s (default), you won't see sequence number repeat until 256s (about 4 min). So, there is a possibility that if your target takes long to respond from time to time, old replies may show up with the replies from the current probe cycle (hence duplicate replies).
Usually duplicate replies themselves are not the problem. They just indicate there is something wrong in the network or at the node.
(Taking a look at your config will also help debug this issue further.)