Probably a corollary of the fact that most usecases can be served by an RDBMS running on a decently specced machine, or on different machines by sharding intelligently. The number of usecases for actual distributed DBs and transactions is probably not that high.