Currently I have clients on both the East Coast / West Coast of the US and my application / database servers reside in Western Canada. My client application is Silverlight based and some of my clients have extremely strict desktop requirements. I'm not even sure if I'll be able to extend Isolated Storage by very much and installing anything locally will never happen. Also some of them have horrendous upload limits.
The Amazon SimpleDB is essentially free for < 1GB and I would never ever have that much data so IU have been toying with the idea of using it like a cache for simple data. Much of my application logic in the database is based around metadata that doesn't change very often so I can almost 'stage' a published version which could be pushed out to SimpleDB. My consuming requirements for that data is never anything more than select * and it is all readonly.
I was also contemplating using it for an auto-save functionality that currently is not stable with the roundtrip via the internet. One of my core features is a data entry module in ERV format so my data is essentially key-value (int/string/date) which means I could rapidly push values to the 'closest' client amazon server and then pull it back down to the core server in the background.
I think for some implementations it could be extremely useful. I did a bit of googling and only MindScape's LightSpeed ORM is supporting it. (Maybe a sign...).
Anyhow it has such limited capabilities that I doubt it would be much effort to implement but it would be nice to have a robust implementaiton intergrated within OpenAccess.
*UPDATE* NetFlix has gone very heavy into the use of SimpleDB. There is some excellent information here: http://practicalcloudcomputing.com
in the guys blog (tagged as SimpleDB).