The test makes the incorrect assumptions that when a chain of 3 transactions involving the same scripthash is broadcast, the Electrum server will send 3 notifications.
In reality the server will queue 3 notifications, but when it computes the status of the indexed transactions for the first notification Chronik may already have indexed 2 or all 3 transactions. The Electrum server only actually sends the notification if the status changes, so if the first notificaton already accounts for the second transaction, it will not send out the second notification.
Fix this by actually waiting for the previous notification before sending the next transaction.