We're considering launching a new check,
It'd be appreciated if you could take a moment to look for false negatives, and let us know.
I found 15 relays where check2 told me that I'm not using Tor.
Very good idea! :) Did you also run the same test against check.torproject.org?
Keep in mind that while it may be frustrating to some users config tests to find that they are temporarily using a false negative exit... this 'multihoming' could, and should in fact, be considered a feature since it keeps some exits out of the reach of anti-tor blocklists that simply parse all the IP's out of the descriptor set. Similar to how obfs and other design features keep Tor out of the reach of anti-tor entities, making an anti to do a little work to be an anti is a good thing. Thus having a good number of false negatives around can allow you to reach places you couldn't otherwise reach due to that type of blocking. That may cause philosophical stress to some places that indiscriminantly [1] block Tor IP's by default, but also keep in mind that Tor is dual use and such a policy is rather blind to that by definition... as would be indiscriminantly blocking all coffee shop IP's by default. So if the point of 'looking for false negatives' is to do something, like attempt to convince the operators to not multihome... the better thing to do would be to add regular false negative tests to the backend of the check service so that these exits come up positive therein.
[1] Within a day ago I posted a link to a slashdot discussion of one particular instance of this.