Testnet historical transaction unavailable via tx-block but recoverable via sui replay

Technical Support
,
1

Hi Sui team,

I’m Reece, founder of Nerava Protocol. During upgrade-resilience and historical-state testing on Sui Testnet, we observed a distinction between current object-state availability, historical transaction retrieval through sui client tx-block, and historical execution recovery through sui replay.

Environment

  • Network: Sui Testnet
  • Initial Sui CLI: sui 1.79.0-46f18562f1f5
  • Observation first recorded: 14 September 2026
  • Later CLI/toolchain: sui 1.80.1-671ba71e69c7

Initial observation

Several historical transaction digests referenced by surviving/current Nerava Testnet state returned not-found responses using:

​sui client tx-block <DIGEST> --json

Examples:

  • HWhiKnos5MCVssSjfsUXsjKe8afD8csBT5x4KhNWnqxN
  • 6HZ5kmL4Z12YU7dNijMAcKxe3sHKrEcj4avi4oLwPABH
  • D4aAw3Qi1SHBk8bYbz8RzcGZCouhcBianK4tPKs6qayi

A more recent control transaction queried through the same command/endpoint was successfully returned:

8EvwPcwgDppeAmCys41kC9ir7XfKUKaSWyLQXNRDbgtU

Its effects referenced the three older digests above among its dependencies.

Importantly, the relevant canonical historical objects themselves remained directly readable in current Testnet state.

Replay observation

We later tested two known historical Allocation-to-Vault Evidence transactions.

For example:

​sui replay --digest AiEbPzRFLyea3fTP2wHMQ4AZo9LfiFHVb9pn8bjtW1xo --node testnet

Replay succeeded and recovered structured historical execution information, including creation of shared object:

0x5917a89f5dd46e80fd998d2591800d9dc6608afe8ed9d01944a2f8b5adb5201f

A direct current-state read independently confirmed that exact object still exists and retains:

prevTx = AiEbPzRFLyea3fTP2wHMQ4AZo9LfiFHVb9pn8bjtW1xo

We reproduced the same distinction with:

HWhiKnos5MCVssSjfsUXsjKe8afD8csBT5x4KhNWnqxN

Standard transaction-block retrieval reported it as not found, while:

​sui replay --digest HWhiKnos5MCVssSjfsUXsjKe8afD8csBT5x4KhNWnqxN --node testnet

successfully recovered historical execution/effects.

What we’re trying to understand

We’re not presenting this as evidence of state loss, consensus failure, a security vulnerability, or a confirmed Sui protocol defect.

The wider investigation also found separate historical-discovery weaknesses in Nerava’s own architecture. We’ve remediated those independently and no longer depend on historical transaction reconstruction for canonical discovery.

We’d appreciate clarification on:

  1. Is this distinction between live state and standard historical transaction availability expected on Testnet?
  2. Is it expected for sui replay to recover execution information for a digest that sui client tx-block reports as not found?
  3. At a high level, what data source/retention mechanism enables replay in this situation?
  4. What is the recommended mechanism for applications that need historical transaction effects for provenance, auditing, migration or recovery?
  5. Are there relevant differences in historical-data availability/retention expectations between Testnet and Mainnet?

I’ve prepared a more detailed technical report containing the full evidence, controls and object relationships and can provide it if useful.

Thanks,
Reece
Nerava Protocol