The data-availability question
For a rollup, data availability asks whether information needed to reconstruct blocks or verify transaction state is published and accessible. Correct execution and later access to data are separate properties. If data is withheld, users may struggle to verify state or recover assets.
Separate publication from execution
Data-availability sampling lets participants request random pieces instead of downloading the entire dataset, using observations to estimate whether it has been published. More independent samples can increase the chance of detecting missing data, but results depend on sample count, network connectivity, and assumptions about honest participants.
How sampling works
In project documentation, identify where data is posted, the sampling protocol, retention period, validator requirements, and data-recovery or exit path. Separate the security role of the base layer from any external data service. The phrase ‘uses sampling’ alone does not establish a particular availability guarantee.
A hypothetical verification
If an attacker hides some pieces of a hypothetical data block, one can ask how likely sampling is to discover the missing region. That reasoning still assumes participants are sufficiently connected and receive responses within the relevant period. Read mathematical guarantees alongside real network conditions.
Assumptions about participants and networks
Record sample size, repetition, assumptions about malicious participants, publication proofs, and failure handling from technical documentation or research. Compare the implementation and network parameters with the description in current release notes.
Check recovery and exit paths
Sampling is not a universal defense against all data loss, censorship, or client failures. Also review upgrade authority at the data service, incident handling, and whether exits remain possible during base-chain congestion. A practical test is whether users can verify state and access assets during an outage.