the minute you have a client (browser, e.g.) and a server you're doing a distributed system and you should be thinking a little bit about edge cases like loss of connection, incomplete tx. a lot of the goto protocols (tcp, http, even stuff like s3) are built with the complexities of distributed systems in mind so for most basic cases, a little thought goes a long way. but you get weird shit happening all the time (that may be tolerable) if you don't put any effort into it.