How? The GOP gained nothing in 2006, and Dems gained nothing in 2010 and 2014? If anything these days, the out-party losing a seat to the in-party is the unusual thing.
I'd have to look at the exact maps at the time, but 2006 wasn't as top-heavy for Democrats. Now they hold a lot of seats and a number of which are in heavily Republican states. Just the sheer number of Democratic seats up and where they are matters more than those particular elections. There is more exposure and more risk. The "out" party isn't guaranteed to win every Senate race in this kind of situation, they just have a really big advantage, and that advantage is reduced depending on the partisan lean of a state.
The average used by 538 when they wrote an article about this was using election cycles that go back a lot further, so the new average these days is probably closer to 100% retention rate than the previous 95% or so. But again, this is arguably one of if not thee worst map either party has faced since the 17th amendment was ratified, so it's not quite the same. Holding every seat and making new gains is tough but obviously not impossible.
I still think Democrats are more likely to do that than not, but I'd be lying if I wasn't a bit nervous about North Dakota. The lack of data makes it really hard to know what is going on.
Remember that 538's fundamentals in ND would give Democrats the advantage, and it has pointed out that in the past 30 years, most Senate races have seen polls shift toward fundamentals during the final months before Election Day. 538 also states that since ND doesn't have a lot of high-quality polling, the fundamentals may have more influence over the final result than places with plenty of high-quality polling.
I'll continue to monitor the polls over the next month, but let me also remind everyone that back in 2012, this race was judged to be Tilt or Lean Republican by several reputable election prediction outlets during the final week before Election Day, but the actual result went the other way.
