Polymarket and Kalshi each resolve through layered states — oracle adjudication, adapter consumption and redemption on one side; venue determination, public finalization and REST settlement on the other — so 'settled' is not a shared timestamp. The author proposes a comparison certificate: a cross-venue duration holds only when payoff semantics, endpoints, calendar support, censoring and observation grades align. Evidence spans 108,638 exact-linked Polymarket conditions and 7,357,576 Kalshi MVE endpoints, yet a matched contrast awaits blind adjudication and a semantic registry.

Public evidence can make an outcome knowable before Kalshi records its own determination, and this paper builds a clock for that gap: first decidability, the earliest moment the rules and a source release map to one outcome, and stable decidability, the earliest moment after which that mapping never changes again. The empirical result is an observability finding rather than a claim about markets: across 152,694 historical Kalshi tickers the author recovered current rule text for all 25 pilot identities but no exact or bounded historical rule version and no exact source-release object, leaving zero historically eligible events. A prospective monitoring stack covering three official source programmes across 25 markets is now enrolling.

A Kalshi market doesn't finalize at one clock, and this registered 7+7 study separates the endpoint from the path. Exact public finalization is observed for 7,357,576 of 7,611,594 MVE market objects and 71,657 of 152,694 ordinary markets — yet only 126,806 MVE and 70,979 ordinary tickers have an exact determination-to-endpoint pair. MVE pairs average 9.676 seconds; ordinary pairs average 322.50 seconds and stretch to 7,133. Settlement speed is not one scalar until endpoint, risk set and observation surface are fixed.

Polymarket markets don't resolve at one moment. They move through a chain: rules written, oracle asked, answer proposed, disputed or not, answer final, payout recorded, money collected. Each step takes its own time. Asking the oracle can start days before a market is answerable; on newer routes the median wait to a first answer is 2 to 8.6 days, while a post-dispute reset resolves in 5 to 48 minutes. Rules also keep changing mid-resolution. Disputes are rare, about 1% of proposals. The lesson for leverage builders: don't use a countdown timer, treat each state separately. Part I of two.