(see T127390: Meet with Security and Ops about a long-term productionization plan for background)
Since a fully production dashboard is still a long way off, we need to run the application with as much stability and performance as can be afforded by Labs infrastructure. The question came up in our meeting about whether it's possible to get read/write access from Labs to a single database on bare metal hardware (or at least something with a performant/non-ephemeral volume), or whether we should instead set up DB replication across a handful (2+) of Labs instances to ensure better availability and balance reads.