Take a simple model of liquidity. I can sell at a good price only if I think you — you in the broad aggregate and collective sense — won’t. (Remember: "Liquidity is only there when you don’t need it") In other words, my motive in selling has to be idiosyncratic.
Now say some common knowledge comes to this market. No one can sell in response to this common knowledge. Everyone has it, by definition, so it’s not idiosyncratic. Transparency, by the way, is simply more common knowledge and that is, with respect to liquidity, the problem in the first place.
Let’s say the new common knowledge is "this asset class isn’t as liquid as we used to think." Ideally price should fall but how much? If selling is only scattered the market never learns the shape or exact location of the new demand curve. Furthermore the selling you observe only tells you "how good is the market at responding to this knowledge shock" and not "what was the initial liquidity downgrading in the first place." Convergence, today, appears to be problematic.
Does herd behavior, combined with agency problems, make things worse?
Is it the standard story that everyone is afraid of the other trader’s knowledge? Or can liquidity crises become more acute in a hyper-informed world? We like to think: "market — trade — liquidity — good, etc.", forgetting the Glosten-Milgrom point that liquidity often rests upon the presence of fools. Informing the fools eliminates one business cycle problem but creates another.
Addendum: Felix Salmon adds excellent commentary.