The blockchain size is already an issue, and running a full node in areas where bandwidth is capped is practically impossible. Satellite ISP, for example, has a ~40GB cap in the US. The blockchain will exceed that size in around a year, possibly less - and that doesn't include the much higher amount of bandwidth consumed relaying transactions and blocks. It's increasingly the case that you can't run a full node unless you live in a city in a country where bandwidth caps aren't allowed by consumers (or maybe the government).
Im already in this boat. I live in the sticks with wireless EVDO (think 3G) for internet.. no cable or DSL out here. its currently capped at 10 GB per month. luckily Im grandfathered in from 2007 with unlimited bandwidth but others on EVDO or satellite may be out of luck, they wouldnt even be able to pull the current blockchain in one shot.
Exact same issue here. 0-100kb/s generally, with some days having odd 300-400kb/s times. Luckily, near a highway. 4G is supposed to be coming around fairly soon. Though... I've been hearing that for about a year. I don't think many people realize just how many people in the rural US (which is most of the US) are stuck with dial-up, satellite, or 3G (*maybe* 4G, though it's about as likely there's only "2.5G" or just no mobile data solution) if they're a bit more up-to-date on the whole "technology craze." Dial-up's no longer usable in this media-heavy Internet, and satellite's too expensive for most (on top of being extremely restrictive), so there're tens of millions in the US without Internet at all. Being in a rural area doesn't just make Bitcoin difficult to use, but the whole Internet is slow and frequently returning time-out errors. Most people just aren't going to deal with that. The Internet obviously didn't die from becoming increasingly bandwidth-demanding, but web developers and designers are very conscious about minimizing data required. The increase rate of bandwidth-demand is many factors lower than that of Bitcoin.
I don't think that's a Bitcoin-killing problem on its own, but I think Bitcoin's way ahead of its time, and I seriously doubt infrastructure's ability to catch up (in most places, it's already heavily-strained). FiOS has been around for, what, ten years? It's deployed only in the largest of cities, and many large cities still lack it. Even then, we're talking 300mb/s absolute max. Max speeds actually offered are ~37.5mb/s down. With a 200GB blockchain, we're talking about an absolute minimum download speed of ~1 1/2 hours. By that time, FiOS will probably be available in all major cities. By the time it's available in minor cities, we're probably talking a 20TB blockchain (unless there's a push against increasing the block size limit, instead finally allowing transaction fees to increase). At that point, the absolute minimum download speed is almost a week. A cable connection would be Hellish (>18 days on the fastest of cable connections, which'll probably be more common by that time), and the fastest DSL connections simply couldn't keep up. The only reasonable way to transfer the blockchain would be physically mailing external hard drives.
No matter what % of data can be pruned - unless that's >95% or the % increases with time up to 99.99999%+, full nodes are going to become more centralized, and they're going to be only in major cities, and they'll be extremely expensive to maintain, which isn't something Bitcoin's prepared to deal with. Bitcoin's been exploding in popularity, but the number of nodes isn't keeping up. Proportional to price, the number of full nodes on the network is spiraling downward, though it's still currently in pretty darn good shape (~190K active full nodes, which has slipped from ~250K active full nodes at the end of November). In September (keeping in mind, lite clients were just as well-known back them), price was ~$150 with ~100K full nodes - so ~667 nodes per $ in price. With price at ~$900, we have those 190K full nodes - so ~211 nodes per $ in price.
ETA: I'm not trying to suggest there's currently any immediate threat to the integrity of Bitcoin. Even 10K nodes wouldn't be particularly worrying, unless the trend continues downward from that point. I think the situation's definitely worth monitoring, though.