There's quite a bit of talk about the potential of DNA synthesis; everything from simple oligo synthesis to whole genome. For all that talk, we've seen custom oligo prices and turn-around time come down, but we've yet to get long writes easily, which would enhance engineering efforts substantially.

Really, what methods and progress is there in DNA synthesis that looks viable? Are we far from commodity megabase synthesis, or do we not need it?

DNA synthesis progress has largely been stagnant this decade - not much has been going on in academia to improve fundamental phosphoramidite chemistry. There've been quite a few papers published (mainly by Church/his offspring, e.g. Tian) to take advantage of cheap (but error prone) multiplexed oligo synthesis via microarrays using enzymatic error correction/NGS/hybridization/amplification schemes.

Otherwise, progress in DNA synthesis has been limited to process engineering, as well as informatics tools (e.g. codon optimization, secondary structure analysis). As a result, the only trend in DNA synthesis this decade has been relocation to China (pretty similar to the current solar PV market, actually) to save on labor costs.