The error you're seeing happens when the covering grid isn't properly filled with data sampled from the original AMR dataset. Unfortunately, It looks like the covering grid isn't being generated correctly in the current version of yt. I'm no expert in this part of the codebase so I don't think I can help out much in debugging the issue.
There's no need to register a bitbucket account to report an issue, but if you do you will get e-mails when there are updates and when the issue is fixed.
It would ease debugging and increase the probability that your issue will be fixed if you share a script that runs successfully in the old version of yt and unsuccessfully in the current develoment version, as well as the mercurial changeset hash (obtained by running `yt instinfo` from the command line) for the old yt version. I believe Matt has some Nyx test datasets, but if your script acts on a smallish (< 1 GB) dataset that you can share (via e.g. dropbox, google drive, or a personal website), that would also ease debugging.
Again, sorry to hear about the regression, hopefully we'll be able to figure out the root cause of the issue here.