Would it be possible to grab a pcap that recreates this behavior? Certainly not the correct behavior and it sounds like you've thought through the potential issues pretty thoroughly already and I agree with your thoughts. We might be at the point of just needing the PCAP to see what's causing it.
It turned out that there is an issue in our network and we are in fact getting duplicated dns packets on the span port...
So bro sees only one dns session in dns.log (and only one uid in conn.log), but the event dns_request is raised more than once and hence we get multiple intel matches.
The usual case for this is that you are tapping the same traffic twice. If you look up the CP1BZx1QgzdPpfEyda connection in the conn.log and look at orig_pkts and resp_pkts you should see 1 and 1. If you see 2,2 or 2,1 then you are seeing duplicate packets.
I am in fact seeing 2,2 or 2,0 as orig_pkts and resp_pkts. And I confirmed this with tcpdump. So I believe it is an issue with the network we are tapping as I see these duplicated packets only for dns.
Possibly, but you may have duplicates everywhere. The tcp reassembler can use the sequence numbers to avoid analyzing the same traffic twice, but UDP doesn’t have anything like that. DNS is just the place you tend to notice the duplicate traffic the most.