Polygon Chain has activated Ithaca on mainnet at block 50,185,000, a protocol upgrade built around a less glamorous but more important promise: payments should keep moving when the machinery behind them has a bad moment. In its release post, Polygon says the update adds automatic failover for stalled block producers, new limits for oversized transactions and clearer operational visibility for node operators.
The update is the fourth Polygon Chain mainnet upgrade in less than five months, following Lisovo, Giugliano and Zurich. The sequence shows a network team spending its release cadence on the settlement layer itself rather than packaging every change as a user-facing feature or token event.
What changes when a producer stalls
Before Ithaca, Polygon says a problem with the node responsible for producing blocks could leave the network stalled for roughly 15 minutes while operators recovered it. The new failover rotates in a healthy producer automatically, giving transactions a route through the interruption instead of waiting for a manual response.
That matters most for the workloads Polygon is targeting: checkout flows, treasury movements and other payments where a delayed confirmation becomes a customer-service problem. Reliability is measured in the missed edge cases, not the average transaction that clears on a quiet day.
Guardrails for the operators
Ithaca also blocks oversized transactions before they can drag the network off pace and gives node operators earlier signals when something is wrong. The changes are aimed at keeping performance predictable under load, a different goal from simply raising a headline throughput number.
The upgrade has a concrete compatibility requirement. Polygon operators must run Heimdall v0.10.0, and the release post says mixed versions across the activation height are unsupported. That makes the operator rollout part of the upgrade itself: the code is live, but the operational benefit depends on the infrastructure around it moving together.
The next useful evidence will come from production behavior during real payment demand: recovery time after a producer fault, the frequency of rejected oversized transactions and whether operators can diagnose incidents without coordination bottlenecks.
