The latency would be incurred several times when loading a single wall. As an experiment, take a moderately complex web app and deploy it in a different continent from the database it connects to.

If you're at the point of storing data on different continents and being aware where it is stored, I'm not sure what is stopping you from batching your trans-continent queries so that you only hit the trans-continent latency once.

Are you suggesting hosting a web app in the USA and its database in Europe makes no difference?

There's also the problem of bandwidth (if you can't cache data locally).

No, but I would suggest that it's not a dealbreaker.

Again, bandwidth pales in comparison to megabytes of Javascript and images getting pushed all over the place. On my nearly-blank test account loading Facebook.com fetches ~4 MB, including ~3 MB of Javascript with instructive names like https://static.xx.fbcdn.net/rsrc.php/v2/y0/r/64jGxSfxJ36.js and https://static.xx.fbcdn.net/rsrc.php/v2/yp/r/K6ojr4ngQRr.js

In face of that, suggesting having to store some data 0.05 s away is problematic is a bit of a waffle