hmm.. FPPS doesn’t really make sense for a DATUM pool. The pool operator doesn’t control the block template (miners do), so fee attribution is fuzzy and the variance/treasury risk sits on the pool. FPPS works today mostly because legacy pools own the template (and often a private mempool). With DATUM, per-block PPLNS fits better: pay on confirmed coinbase using a last-B-blocks (or N-difficulty) window.
Building such a pool is more grunt work, totally doable. Turning this PoC into a realistic PPLNS/(or similar) DATUM pool is mostly plumbing and ops — not research. This was the whole point about my post. If pool owners are motivated for decentralized mining where miners control template then DATUM is just ready to go.
DB-wise: keep it boring. Postgres for users/rounds/credits/payouts. If you want raw share lineage, add Timescale for a shares hypertable; otherwise store per-height aggregates only.
