We rely heavily on S3 storage, including in our production use, but there are no regularly-run integration tests for the functionality.
We do already have a RestAssured test class (test/java/edu/harvard/iq/dataverse/api/S3AccessIT.java) with one test, written by @bsilverstein (hi!). We are not regularly running it on jenkins.dataverse.org.
S3 needs to be configured on the test bed installation for the test to work. But since our new jenkins infrastructure already depends on utilizing AWS, this is not an issue. We would just need to modify our scripts to copy the AWS credentials to the target EC2 instance and configure an S3 file store there.
So I propose we do that; plus extend the S3 test class above, for some more extensive testing and to test the functionality that's been added since it was written. (for example, as it is now, the test relies on having S3 configured as the default storage driver. We can instead make it create a new dataverse, set S3 as the storage for that specific dataverse, using the new multistore functionality, and proceed testing it)
S3 testing was specifically requested in #5068. But I'm opening a dedicated issue for it since there were other things being discussed there.