Litecoin Core fee estimation: block history and wallet fallback
Release: fee-estimator-20261026-v1
Editorial/source review: 26 October 2026, Europe/Warsaw.
Published execution: 25 October 2026, 23:32:22–23:33:02 UTC (26 October in Warsaw).
Software: official Litecoin Core 0.21.5.8 Windows x86_64 executables.

QUESTION AND RESULT
A new node was asked for economical/conservative estimates at targets 2,6,12.
After generating 101 and then 301 coinbase-only regtest blocks, every requested
estimate still omitted feerate, reported an insufficient-data error, and
returned blocks:0. Each of all 301 blocks was checked via getblock.
After 101 and 301 blocks, initialblockdownload was false, headers equalled blocks,
and recorded tip/current-clock inputs met the pinned source currentness gate.
No eligible mempool payment observations were supplied in this experiment.
This is not a claim that every absent quote has the same cause.

Two construction attempts with fallback disabled failed with code -4. After
restarting that same disposable node at height 301, three explicit configurations
produced signed 141 vB transparent candidates:
- Fallback 2 litoshi/vB, wallet/relay minimum 1: 282 litoshis.
- Fallback 0.01 litoshi/vB, wallet/relay minimum 1: 141 litoshis.
- Fallback 2 litoshi/vB, wallet minimum 1, relay minimum 10: 1410 litoshis.
All three passed testmempoolaccept. They were never broadcast or mined.
All numbers are laboratory settings/results, not live prices or recommendations.
The RPC estimator remained unavailable in all three construction cases.

The exported run passed 55 experiment assertions. An earlier fresh-chain run
matched block counts and the three fees; repeat-check.json gives its scope.
The repeat was internal, not an independent external attestation.

CAUSAL BOUNDARIES
Fresh height 0 was in IBD; that is retained rather than hidden. Later checkpoints
were not. Source checks cover MAX_FEE_ESTIMATION_TIP_AGE, IsCurrentForFeeEstimation,
validForFeeEstimation, estimator processTransaction/processBlock and first recorded
height. We did not inspect private estimator objects through a debugger.
No training transaction was broadcast, no fee-estimator success was demonstrated,
and no claim is made about the number of real-world samples needed for a quote.
The stronger relay case changes static local policy; it does not fill the mempool
or measure the rolling floor caused by pressure/eviction. A normal restart may
read retained fee_estimates.dat; a new directory is a different condition.

ISOLATION
Every run creates a fresh unique datadir and disposable wallet. The script never
accepts an existing datadir, wallet, configuration, keys or seed. RPC binds to
loopback. Peer connections/listening, DNS discovery, UPnP and onion listening are
disabled. Resource settings are explicit. All peer counts remained zero.
Only new regtest keys and valueless test-chain coins were used. No real payment,
mainnet wallet, public RPC, hardware/mobile wallet or MWEB flow was exercised.
The started process was cleanly stopped before its temporary directory was
removed. Cleanup checks never target a previously existing directory.
No elapsed duration is a confirmation or node-performance benchmark.

REPRODUCE
Verify official Core executables against the project release independently.
  python fee-estimator-test.py.txt /path/litecoind /path/litecoin-cli new-output
Windows: quote full paths to the .exe files. Python 3 standard library required.
The output directory must be new. An optional --scratch-parent chooses a parent
for a newly created temporary lab; it does not accept an existing node directory.
Binaries and signing-key provenance are documented in binary-provenance.json.
Fresh wallets, addresses, identifiers, signatures and timestamps vary. Signed
size can vary slightly; the published 141 vB values describe the recorded run.
The script does not download software or export signing secrets/cookies.

VERIFY EXPORTS WITHOUT CORE OR NETWORK
  python verify-results.py.txt /path/to/package
The verifier reconstructs ordinary transparent TXIDs, weights, virtual sizes and
input-minus-output fees using included previous transaction bytes. It reconciles
RPC/candidate records, 301 block rows, gates and assertions. Optional package
checksums are verified too. It does not validate signatures/scripts, independently
attest that the supplied trace was executed, or parse/validate MWEB transactions.

FILES
results.json: selected checkpoints, construction attempts and assertions.
rpc-trace.json: public RPC calls/results for this experiment, without auth headers.
blocks.csv: every generated block's height/hash/time/transaction count.
checkpoints.csv: synchronization and estimator-currentness input summary.
wallet-candidates.csv: explicit configuration and observed construction fees.
cleanup.json: own-process/disposable-lab cleanup summary.
repeat-check.json: comparison with one earlier internally repeated run.
fee-estimator-test.py.txt: complete experiment source.
verify-results.py.txt: independent export/byte-arithmetic verifier source.
sources.json: versioned primary source mapping and source capture hashes.
binary-provenance.json: verified release/download metadata, no executable payload.
fee-estimate-checklist.txt: blank diagnostic aid, no secrets requested.
SHA256SUMS.txt: file integrity hashes; not a third-party review or signature.

All addresses and transaction identifiers here belong to isolated regtest chains.
Do not send real Litecoin to them or expect mainnet explorers to find them.
No RPC cookie/password, seed/private key, real wallet file, host username or
absolute workstation path is included. Runtime binaries/private logs are absent.
Research, code and drafting used AI assistance. No independent external review,
mainnet fee quote, timing study or service-availability guarantee is claimed.