when looking at the ledgers endpoints for dashboard-v2 and dashboard v1 (in prod now) I noticed the tx and op count data are slightly different. I'm guessing it's due to using data in memory as part of the sum equation in the updateCache method, in which case a mutex may help (maybe async-mutex). But some further investigation is probably needed.
Not high priority, they're all fairly close (~95% or higher), and this data is mostly used for visually comparing days in the UI, so data doesn't need to be exact.