This reminds me of a time when someone wanted to send me 17 5Gb docker images of data, that later needed to be merged.

We made them produce a tar.gz instead. I've never understood why people put data in containers, given the alternatives that are meant for managing data

Container file system layers are based off of tar files; some folks prefer containers for data management due to the tooling available vs tar/gzip. You can compose using a Dockerfile, version with hashes, label, and push/pull/transfer with registries.

I think there are benefits in using containers for data management when properly implemented.

I do it for development purposes of heavily data (state) dependant applications to roll back or jump to various specific scenarios very quickly

I agree. I would never put data in a container in production, but for development purposes it can be helpful.