Say my aggregate vega is -100 for a position consisting of several options, all with the same underlying. AFAIK vega measures sensitivity to IV changes. Only problem is IV has no aggregate measure accross a portfolio, so how is aggregate vega useful if you have no aggregate IV measure? What am I missing here?

It's like aggregate delta. You buy 10 stocks (100,000 each) and say you are long 1MM worth of stocks. It's not 100% accurate but it's a way to describe your portfolio. Also, like delta, vol tends to be correlated with eachother. EDIT: Misread. Didn't see the same underlying.

So Vega measures P/L based on a 1% parallell move either up or down in IV accross the entire volatility surface?

Makes sense. But how can I be sure the surface moves in a parallell fashion? What about skew and convexity risk? How can I measure that? Or maybe those are negligible in vanilla options?

The slope can decrease on the down and outs on an index selloff, so you can't measure parallel. The risk would be fairly straightforward to sell deep otm gamma if the slope was static. You can take the 15D puts less the ATM vol/2 and use that as your mid-point skew figure on the strip. Adjust that figure +/- 100bp and you'll have a curve of 50D, 15D on the downside and mid-curve vols. It's very crude. 3rd and > moments are likely overkill unless you're concentrated in a leveraged skew book. PM me if you need specifics.