4.3.0

Added

• IntervalAwareMetricAggregatorInterface: an optional interface an aggregator can implement to declare a minimum number of seconds between runs. AggregateMetricsCron skips such an aggregator until that interval has passed, using the Magento cache to remember the last successful run. Aggregators that do not implement it keep their existing per-cron-tick behaviour, so this is additive for third-party aggregators.

Fixed

• OrderAmountAggregator, OrderCountAggregator, OrderItemAmountAggregator, OrderItemCountAggregator: all four still run an unbounded full-table scan over sales_order / sales_order_item on every cron tick even after the N+1 fix in 4.2.1, and are the largest contributors to the DB load reported in #69. They now implement IntervalAwareMetricAggregatorInterface and are throttled to once every 5 minutes instead of every minute, while every other (cheap) aggregator keeps its existing per-minute schedule.
• AggregateMetricsCron: a cache-backend exception while checking/marking an interval-throttled aggregator's last run no longer aborts the rest of that cron tick's aggregator loop, and an aggregator whose aggregate() returns false (a soft failure) is no longer falsely marked as a successful run - it retries on the next tick instead of being suppressed for 5 minutes.

Upgrade notes

• AggregateMetricsCron gained a Magento\Framework\App\CacheInterface constructor argument and is injected as a proxy in src/etc/di.xml, so run bin/magento setup:di:compile after updating.
• The throttle marker is stored in the Magento cache without tags. bin/magento cache:clean does not clear it; bin/magento cache:flush does. To force a throttled metric to recompute immediately, use bin/magento run_as_root:metric:collect --only=<metric_code>, which ignores the throttle by design.