Coming from Timescale, the proposal is not to use the best kind of database for a workload, but to use one database but then multiple extensions for workloads t

Coming from Timescale, the proposal is not to use the best kind of database for a workload, but to use one database but then multiple extensions for workloads that Postgres may or not be the best fit, moving the multiple layer one level down. And maybe not getting the best performance (database people are specially sensitive on that topic, for heavy loads at least), that may or not introduce incompatibilities, inconsistency or compromise stability. Shaving complexity is a good goal, but is partially negated by adding it again closer to the core. Anyway, maybe for most (in unique installations, or companies, not amount of servers dedicated) the workload won’t be heavy enough to deserve the extra complexity of handling many different database servers.