Feb 25, 2026

Not only are the fee computations now direct relational queries, but for blocks they are executed concurrently by tx. Because spend and value are both output in the API, this results in two independent calls to summarize inputs and outputs, both summing concurrently over txs in the block. For a single query to just obtain the fees for each tx, or for the block as a whole, this is further optimized to operating in a single concurrent run over all txs. That's not used above but is used to initialize the fee estimation accumulator, which by default collects fee data for all txs in the top 10,000 blocks. I expect this to now be sufficiently fast to allow it to initialize from the store at each server startup (when enabled), so there will be no need to cache fee accumulator state (e.g. as BC does). If each block takes 1ms then a 1000 block fee accumulator will be initialized just 1sec after startup. For a cold stare this will take a little longer, but should still be sufficiently fast. And there's no need to block startup for this, it can just give service callers a busy response until ready.