Increasing block size
I've been keeping a close eye on the average block size. It has been consistently over the 25% mark for the last week, and shows no signs of relenting. Plenty of new users have run into low bandwidth issues, which is an absolutely unacceptable user experience. I'm glad to see 
Regarding scalability issues. I've verified that a 500 GB NVMe or a 500 GB SATA RAID0 based-server is affordable for most witnesses above the rank of #100, and currently this uses less than 30 GB shared-memory + 60 GB block_log. You could store the block_log on a slower SSD or even HDD, leaving plenty of room for growth on the NVMe partition to host shared-memory. Currently, this server uses less than 5% disk I/O, so this is not a bottleneck. Furthermore, I expect AppBase to bring plenty of further scalability improvements. In a nutshell, I don't believe scalability to be an issue at all for any witness above the rank of #100 (and this is a very conservative estimate). Witnesses ranked #50 and above can also store shared-memory in RAM (256 GB servers) for the foreseeable future.
TL;DR - The block size must be increased to sustain the Steem network's exponential growth. Witnesses are paid enough for this not to be a sustainability issue.
Spectre/Meltdown patches cripple performance on Intel systems?
I've discovered that the Spectre/Meltdown patches have a significant impact on disk I/O performance with Steemd on Intel chips using Ubuntu 16.04 LTS. I will consider this anecdotal, however, and request fellow witnesses/developers to report their findings. I'm unable to confirm if AMD Ryzen or EPYC servers suffer a similar hit (as the new servers seem patched already), but in most likelihood they do not. This is because AMD has a near-zero risk to Variant 2 and Variant 3. I shall transition to a Ryzen or EPYC based server this week.
SBD / STEEM
The STEEM pump in early January has been an interesting simulation for how a positive bias would affect SBD. Please note that we only have a month's worth of data, so this is hardly enough evidence to draw conclusions. However, there's a clear trend emerging. There are of course several other factors at play, but there have been two main results - a) Authors enjoy far higher payouts, b) SBD is on a sustained downward trajectory closer to its intended USD 1.0 value. There was one irrational pump for SBD mid-January, and a significant sell-off earlier this week, but even correcting for these anomalies, leaves a clear trend forming. Thankfully, the conversions have been disabled in UI through this turbulent period.
To this effect, I continue publishing a bias. Since the higher STEEM price has effectively had the same impact as a positive bias in stabilizing SBD, I have now reduced the bias to 200%. I fully expect SBD to continue its downward trend, barring any other errant pumps. (Which, of course, could happen at any time.)
However, given the rapid pace at which SBDs continue to tumble, I'm concerned about sustaining that USD 1 target, in the event of an exponential rise in STEEM price. I plan to publish negative biases in case SBD dips significantly below USD 1, and set a few BTCs worth of SBD buy orders at the exchanges. Things can change very fast in the crypto market, so witnesses should be well prepared for this eventuality.
Leaving Steemit.chat
I'll no longer be on Steemit.chat. I have requested the (existing) admins to delete my account. I have no idea who the new owners are, and a preliminary investigation doesn't inspire much confidence. This is a personal choice, of course, and I elect to err on the side of caution. If the new owners of Steemit.chat prove to be trustworthy stewards of the services, I'll be back later in the year. Till then, you can find me on Discord 