You say that as if it solves all problems (e.g. fast reads) at the same time.

You could very easily build a link-list type datastore in SQL and use a CTE to build the list, but if your main use-case is viewing the list rather than updating the order, then it's a slower compromise than something like the solutions OP was discussing.

Would you elaborate? Using a graph database would likely include tradeoffs in data integrity (which generally needs to be handled by application code) guarantees and performance (due to indexing). If you have specifics of how you would implement this with the same guarantees and performance in a graph database, please share. Graph databases certainly have a place, but they're not cure all (just as RDBMSs are not).