In their data modelling pages they mention you should break up rows into separate keys per column. Or separate keys per field in a document. This is indeed how many databases model rows on a distributed kv store. So this might be how they achieved 100TB.

However you still have the issue of any single key-value needing to be in their limit. (But it's not like people typically store enormous blobs in Postgres or MySQL either I think?)

The documentation is woefully out of date, sadly. Despite the code being in active development no one is touching the public docs. Though I don’t know for sure, that limitation was probably written something like 10 years ago.