Unfortunately it is only “permanent” until the next time the TI part goes wacky.
If we had repeatable way to make the TI part fail, we might be able to debug it but I cannot find a way to make the error happen on command. Many of the folks affected by this problem have what I would call “funny” networks with high latency or new types connections, but that could be coincidence.
@mtnscott wrote a tool described here to look for this and other problems:
And a more “permanent” solution is to use a DNS client outside the TI part that @Hootie81 wrote/ported.