The "provocative" post  raises an interesting point. Does publishing feeds with high frequency harm the network?
The idea is if everyone's publishing price feeds very frequently, the cost to manipulate the feed is small -- the attacker can simply dump BTS on exchanges, causing a dip in the price feed, then sell much larger BitUSD holdings into margin calls at artificially high prices. A relatively small manipulation -- only needing to clear out the order book depth immediately available on the exchanges -- can be amplified into a much larger manipulation -- margin calls to force covering at inflated prices by the entire short float.
OTOH if you have a random update interval (with some known expected value), the attack is much more expensive. E.g. if average feed publish interval is 8 hours, then (assuming all delegates are online and honestly reporting prices) it takes 4 hours to update the median feed, so the attacker would have to sustain his dump for 4 hours, making it more expensive.
Also, "reactive" publishing -- publishing after a sudden price movement -- would also have a downside, it would mean a lot of delegates would update at the moment the attack commences.
I think we should consider the possibility that feed updates need to be done more slowly so that traders have time to move funds into / across exchanges and force any dump to break through the all the actively trading capital in the ecosystem, not just whatever happened to be immediately available on the exchange(s) when the dump occurred.
The main cost is that if there's a news event at a single discrete point in time that causes a quick price change, margin positions may not immediately be liquidated and the price would have time to fall further before the feed caught up and the positions started to unwind.