I'm not sure whether the above post should be marked "astroturfing" but it sure reads a little too positive.
454's sequencing technology is a welcomed addition to existing technologies, but don't believe the hype, particularly when the person talking has stock options.
The analysis of genomic sequencing data (metagenomics or otherwise) is highly benefited by large contiguous pieces or ideally whole contiguous genomes. Related to this and more fundemental is the fact that the shorter the pieces of DNA spat out by a machine the harder the problem of assembling them into larger contiguous chunks. This is due in part to the combinatorics of an alphabet made up of only 4 symbols but mainly the fact that genomic DNA contains many repeat structures even in lower organisms.
Without going into detail, it suffices to say that the longer the pieces (or "reads") produced by a sequencing machine, the easier the problem. Add to this the realities of sequencing errors and throw in metagenomics where you may have many organisms with almost the same genome, the problem gets quite hard.
Currently the large sequencing facilties that use 454 machines use them to complement their existing machines which produce 3-10 times longer reads (depending on who's talking). There are in fact papers investigating the ideal ratio of reads produced by new and old technologies.
Another factor to keep in mind is that, although the new high-throughput technologies (454 is the first to market, but not the only player) hold alot of promise, a large part of their appeal was going to be an enormous cost reduction. The problem is, so far that part of the equation hasn't met expectation. They are quite costly to run due to the cost of consumables and those prices are set by the manufacturer.