It seems like you answered your own question,the signal from a few molecules running through an enzyme or a polymerase tend to fall out of synch after a few hundred bases. If an enzyme for sequencing was more rigorously in time step that could help for instance. The machines read traces in four channels with nice bumps for each base. See this article for a nice example. You can see that if there are too many of the same base consecutively it gets hard to tell how many bases there are there. Over time all four traces will start to smear out and you cant tell Adam from Thelma if you take my meaning.
But there are other bottlenecks.
The sequencers currently put out such a high volume of data that the analysis of the meaning of the output data can't be analyzed fast enough. This follows the trend in biotech over the past 12 years or so - more sequence data, micro array data, more mutation data, more genomes than people who can actually use it to understand the biology. There is a bit of an analysis bottleneck now.
So some of these sequencers have greater read lengths, which can make it easier to assemble a sequence. These sequencers generally cost more. For instance if you have a library to sequence a little fungal or algal genome - you will get the answer back in a day or less now. In the form of 1 Tb of reads maybe 50 to 200 bp long. It might take quite a bit of time to put that together into a novel genome sequence, still more to find the genes, build the gene networks from a template of pathways etc. Just imagine a thousand sequencers pumping out day and night and you get the picture I'm trying to paint here.
About cost. Ion Torrent and the new oxford nanopore sequencers are really cheap - \$50k to perhaps \$900 for Oxford Nanopore's USB sequencer. Most other systems cost hundreds of thousands of dollars. Ion torrent and Nanopore have more disposables - you throw away a chip or even the entire sequencer - at a cost of hundreds of dollars a sample.