Walrus doesn't simply replicate data; instead, it uses erasure coding. This approach allows for data reconstruction even if a subset of nodes (between one-third to two-thirds) are offline, providing resilience without relying on replicas. More details on this can be found in the whitepaper. Regarding storage duration and costs, Walrus operates in epochs, with payment being upfront. You can specify for how many epochs data should be stored, and extend the lifetime of your data as needed.

There are no read costs for data users since reads typically go through caches to minimize the load on storage nodes. In terms of managing illegal data uploads, while Walrus can't inherently prevent them due to its decentralized nature, nodes can denylist certain data. If enough nodes denylist a particular data, it will be removed from the system. Node storage providers do have some control over what data they store. They can refuse to store or serve specific blobs, giving them some level of control over the content on their nodes.