They look like more than 75% full right now. That could create problems if you need a transaction confirming quickly. You could probably average the blocks out over a few hours to make it look like the average is below 75% full, but that doesn't help people who need transactions confirming quickly right now. There are times when the average is below 45%, and other times when it's above 90% (which is creating problems).
https://blockchain.info/charts/avg-block-size?showDataPoints=true×pan=30days&daysAverageString=1&scale=0&address=
0.71mb = there's an extra 40% capacity.
Ane people are still paying peanuts for the bulk of transactions (free or near free): http://www.cointape.com/
....with not much "urgency" in their tx confirmation - otherwise they'd bump their fees up by a few satoshis.
On the other hand those that pay more normal fees, get included in 0-1 blocks.
The future is kind-of-accounted for (considering the problems of BTC in scaling, which are much deeper with current technology than a change in block size) through
a) devs working on scaling
b) fees that will eliminate a lot of dust/spam that are now processed when they shouldn't
Blocks getting full is a meaningless factoid without seeing the quality of transactions and fees involved. We could have 2 mb blocks tomorrow morning and someone could activate a script and fill them up for peanuts. He wouldn't be in any hurry of course for inclusion, so he'd pay the absolute minimum he could get away with. After all even if 500kb legit transactions are ahead of him and he takes the last 1.5mb of the block with zero/near zero fees, the effect for an outsider would be that "...oh my gawd blocks are full - even after we went from 1mb to 2mb... Oh shit we need 4mb!!!"
If you see threads back in 2013, you'll see they anticipated that given the size used they were expecting a problem in 2014 or so and some were like, bullshit, it's just a lot of spam by satoshi dice and similar stuff. Those who went with just the quantity of data failed to predict that we'd be still running in 2016 with 40-50% capacity to spare, those who examined the quality of data knew there was nothing to worry about. That's not to say that an increase is not needed. Of course it is. Multiple increases will be needed. But nothing will happen even if blocks are full 100% of the time with junk. We've seen emulation runs with the various "stress tests". All normal txs went through with a fee, spam stayed in the queue. The fullblockalypse that never was.