I've been running a few measurements over the last while and looked at the raw results like:

[{"dst_name":"2001:4860:4860::8844","error":{"socket":"connect failed Network is unreachable"},"from":"fd84:d527:5183:0:a2f3:c1ff:fec4:63a8","fw":4940,"group_id":18903906,"lts":21,"msm_id":18903906,"msm_name":"Tdig","prb_id":13635,"....

"from":"2001:db8:4447:0:eade:27ff:fec9:7134","fw":4940,"group_id":18903906,"lts":7,"msm_id":18903906,"msm_name":"Tdig","prb_id":26725,"...]

I'm curious if/how atlas reports back to probe owners: 

  "Hey, your ipv6 connectivity is implausible/impossible"

since, at least for:
  https://atlas.ripe.net/probes/26725/#!tab-network

and:
  https://atlas.ripe.net/probes/13635/#!tab-network

there's no way their v6 addresses can work... (one is documentation prefix the other is ULA I think?)

I suppose test requestors should check the ipv6 prefix to see if it's at least plausibly correct until some signal back to the owners can be attempted?