Playing NL5 zoom and my up/down swings have been a little large over the last few days. Maybe it's cos I'm bored over xmas and getting involved in too many hands. What's a normal kind of variance in terms of number of BIs, do you think?

Is losing 3 BIs in one hour of two tabling zoom a lot? If you then make 4 BIs over the next two hours?

If you're a typical player (not too loose/aggro), then you should get a standard deviation of around 75bb/100, which basically means that on any 100 hand sample, your actual result will be somewhere in a range of 75bb above or below your true long term winrate. Since winrates are often in the 0-10bb/100 range, the variance is huge. i.e. You might "expect" to win 15bb in your next 300 hands, but it's perfectly possible to (or even likely that you'll) win or lose 2 or 3 buy-ins in a sample of that size. It's only over large sample sizes that trends and averages can be discerned.
I doubt it's any consolation to you, but in a recent sample of 30,000 zoom hands, I had three upswings of over 10 buy-ins, and one hideous downswing of 17 buy-ins (the like of which I've never experienced, as it happened so quickly, in the space of 3 or 4 short sessions). The variance felt insane at the time, but in the grand scheme of things it's completely normal. It's just a statistic. Poker is a game of small edges and high variance. (The games have been particularly wild over the Christmas holidays, due to the splashy play of casual/bad players, so your variance might have been a bit higher than normal).

One thing I would suggest, however, is that if the swings are messing with your mind in any way (starting a thread about the variance suggests that they do) try setting yourself a stop loss of 2 buy-ins. A "sudden" loss of 3 or 4 buy-ins due to normal variance can often extend into a 10 buy-in downswing due to tiltspew.

In my experiences, the most common large downswings I'd face were 20 BIs. During sessions, you will frequently swing around up and down, that is incredibly hard to define specifically over the course of hours as you have tried to do.

Not sure about variance, but the standard deviation according to my calculations is about 1.8708