A remote donor shard may return an InterruptedDueToReplStateChange error response as a result of stepping down or transitioning into rollback. The ReshardingOplogFetcher thread will (incorrectly) stops rescheduling itself when this happens.

try {
    return consume(client, factory, targetShard.get());
} catch (const ExceptionForCat<ErrorCategory::Interruption>&) {
    return false;
} ...

The ReshardingOplogFetcher thread should instead swallow the Interruption exception and consult the cancellation token for whether or not it should reschedule itself.