There are solutions - like what Ethereum is trying by using a block tree, rather than a simple chain - but someone else can probably explain that better.
I don't think the blocktree in Ethereum's configuration really addresses this, as the block times are still one minute (or at least, were when I last talk to Vitalik). Aside from that, extremely fast blocktree implementations (as in the GHOST paper) still requires tx to traverse the network among different nodes and will probably hit bandwidth problems comparable to Bitcoin, if not worse because very large numbers of blocks containing the same tx will be hitting the network at nearly the same time any given second. It's more a solution to the problem of insecurity introduced by large blocks on a slow network.