I am just starting to work with using TransactionScope, I find that there are always unexpected things I run into that take forever to debug.
I figure that having a consolidated list of these would be great for those "weird error" circumstances, plus to expand our knowledge of oddness in the platform.
Some context on how I am going to be using transaction scopes:
2 things off the top of my head:
You want transactions to be lightweight as much as possible, dtc introduces a lot of overhead. You also want transactions to be as short as possible, so introduce them only on the application servers and not on the web server. Make the hop over the network between the application servers and the database as small as possible and as fast as possible, send network traffic between web and app servers over a different connection than between app servers and db, and make the last one a screaming fast, ridiculously short connection.
If you have multiple application servers, you could consider having a single instance of msdtc running on a server (e.g. on the database or on one of the app servers) and use this remotely from all application servers instead of each one running their own, but I don't know what additional benefits this has.