Hashrate is still ~21 TH/s... didn't it go up to around 28 a day or two ago, or did I dream that??
Not dreaming, you can see it on
http://www.asicminercharts.com.
I do not know if it was a test run deploying new kit. I've noticed in the past when the hash rate goes up to a new high it tends to come down again before retracing up again. It looks like it's climbing now.
Variance. I'll try to get time to post a chart which should provide a better indication of hashrate than a daily average.
I'm tracking the solo mining hash rate on my site. I have a chart with 24-hour, 3-day, 7-day and all-time average:
http://runeks.dk/bitcoin/
The 24-hour average did climb, but it was probably just a streak of good luck. The 3-day and 7-day averages are all climbing though. It's clear that ASICMiner has added hashing power to solo mining in the past week. But the 24-hour average (much less the 12- and 6-hour average) aren't reliable estimates of the hash rate.
This is only an estimate of the solo hash rate though. You need to add 7 Thash on top of that from BTCGuild. So a good bet is that they're hashing at around 25 Thash/s in total.
A more interesting figure though, is the coins mined since last dividend payout. This is what's really relevant to shareholders. It's already above the level at the previous pay, Wednesday last week, so it's fairly certain that they've added extra hashing power to solo mining:

Again, these charts are only for solo mining. So add another 40% on top of that from BTCGuild. I'm counting on ASICMiner going 100% solo, so I'm not adding a tracker for BTCGuild. But I'm not sure whether than has been announced or not.
When dividends are paid out, is the address the solo mined coins go into zeroed out? Or in general, can you explain in a little detail the second graph and how you get the info for it? I remember previously when looking at your charts the two lines would diverge... why would they ever diverge? Hell, why have the two lines? Shouldn't they... in a way... show the exact same thing? Or is there something I am missing here that would cause a difference in mined vs. mined per share?
Like what's his name has pointed out twice now, since we know exactly how many coins are mined, people really need to ignore hashrate... and I need to better understand this second chart.