Errors during a flush
Errors during a flush in the Scheduler and timing guide.
One failing effect never wedges the rest. Every queued effect runs; errors that no owner handled (see error routing) are collected and thrown once the queues are empty: the error itself when there is one, an AggregateError when there are several.
When the flush was started by the microtask, there is no caller to throw to: the error is passed to reportError (or, where that does not exist, becomes an unhandled promise rejection).