But what if the error causing a failure to process a message is not in the consumer application but in the message content? Then, the first consumer processing the message will keep processing it until the visibility timeout expires, and then the second one will start processing, and so on. What can we do in this case?

The SQS supports implementing dead letter queues when using standard SQS queues. The dead letter queues are designed to prevent a message with corrupt content being processed into infinity, and allows us to develop the application to detect such messages, dequeue them from the main queue, and queue them up into the dead letter queue. This queue is then used by developers and administrators who can take a look at the message and determine the cause of the processing failure.