Litecoin.watch frozen network analyses, release 2026-10-04-v1 (Europe/Warsaw).
Captured 2026-10-03T23:12:19 through 23:12:47 UTC / 4 October 01:12:19 through 01:12:47 Warsaw. One synced, pruned Litecoin Core 0.21.5.8 mainnet node. This release contains a fixed observation; no new persistent collector or scheduled job was created.
Anchor: block 3189039, hash b53ea8dc31a0622a9473607e217a1d9ca15692a204be5d33b224fc31760da9f5.
601 consecutive headers: 3188439 through 3189039. 144 fee blocks: 3188896 through 3189039. Heights were rechecked against the active chain after capture. Capture times are Unix UTC seconds; header times are miner-declared and are NOT node arrival times. Header coverage is not a calendar-day sample.

FILES AND DEFINITIONS
capture.json records boundary, software, capture times, read-only scope, returned estimates and input hashes. headers.json retains height/hash/parent/time/mediantime/bits/difficulty/chainwork. blocks.json retains selected getblockstats responses, coinbase output values, HogEx count and public kernel fee fields. Full transactions, output addresses, wallet files, peer identities and authentication are not exported.
fee-blocks.csv has 144 block rows. All amount columns ending in _litoshis are integers; 1 LTC = 100000000 litoshis. RPC txs includes coinbase and HogEx; kernels are separate protocol records, not users or invoice counts. Totalfee already includes HogEx's MWEB contribution; kernel fees are subtracted for the ordinary base-chain fee decomposition, never added again. Coinbase may underclaim; our 144 rows observed full allowance. Fees/subsidy are LTC-only and exclude DOGE rewards, electricity, hardware, price conversion and pool payout.
One entry's maximum fee in block 3189026 is 0.2 LTC. This block has no kernel fees and total fees 0.20157609 LTC, below twice that maximum, so only one entry can pay it. No transaction identity, sender, cause or purpose was reconstructed. Leave-one-entry-out calculations remove this entry from the selected fee numerator and entry count; they are sensitivities, not alternative observed chains. RPC median includes HogEx; a median of block medians is not global transaction median. Whole-sample mean uses summed fees divided by summed entries.
hashrate-windows.csv has 3 estimates at the same boundary, using 30/120/576 work contributions and 31/121/577 headers. Core denominator is max(time)-min(time) across each selected inclusive boundary range; numerator is ending chainwork minus starting chainwork. It is not endpoint subtraction by definition, median-time subtraction, nominal target seconds, hardware telemetry or difficulty at a single current point multiplied by every block.
rolling-hashrate.csv has 571 overlapping 30-block and 481 overlapping 120-block estimates. The line figure restricts both to 481 matched ending heights 3188559 through 3189039. Full-series descriptive minima/maxima use their explicitly different supports. These are neither independent observations nor confidence intervals. All saved bits are 19301c71; this does NOT demonstrate unchanged running hashpower. No negative adjacent times or endpoint/extrema discrepancy occurred in this range.
Proof increment uses integer floor(2^256/(target+1)) with compact-target decoding. 601 headers provide 600 checked increments. Hashrate is expected proof work / header time; the node's floating return and our Decimal ratio agree within relative 1e-12. No hardware census, attribution, attack probability or causal conclusion about miners joining/leaving is supported.

OFFLINE REPRODUCTION
Keep the original frozen files, including analyze.py.txt, together. Python 3.10+ standard library only:
python analyze.py.txt /absolute/path/to/package-directory /absolute/path/to/new-analysis-directory
Compare generated results.json with the frozen results.json. 2223 assertions check original input checksums, links and proof increments, 144 block accounting records, numerical estimates and accounting identities. These are consistency/arithmetic checks; they do not validate each serialized block independently, query a second node or establish a representative sample. No network, wallet, funds or miner is needed for offline recalculation.

FRESH READ-ONLY CAPTURE
Requires your own synced Litecoin mainnet node, suitable retained recent block/undo data, local RPC cookie and a new output directory. Never share the cookie, data directory, wallet files or credentials. Save capture-node.py.txt as capture-node.py:
python capture-node.py /absolute/path/to/node-data-directory http://127.0.0.1:19432/ /absolute/path/to/new-output-directory
Choose the loopback port your node already uses; 19432 is our study port, not a claim about Core's default. The collector rejects remote-host URLs, reads only allowlisted diagnostic methods, bounds every response, preserves frozen output, queries 601 headers and 144 block/stat records once, and changes no node settings. Our original run took 28 seconds with nice 15; timing on another machine can differ. A fresh boundary will give different observations. The current analyzer includes release-specific checks and an outlier sensitivity tied to this snapshot; update those assumptions and release identity before analyzing a different capture. It is not a general unattended monitoring tool.

SOURCES, LICENSE AND CHECKSUMS
source-records.json contains pinned primary source URLs, fetch times and hashes, checked 4 October Warsaw / 3 October UTC. Full third-party source captures are not redistributed. Original charts are explanatory SVGs, not screenshots of pool software or photographed mining equipment. Blank TXT worksheets are local records, without account integration.
SHA256SUMS.txt covers each content file except itself/ZIP/archive-sha256. The ZIP includes the checksum file; archive-sha256.txt identifies the ZIP separately. Input hashes protect the byte identity used by the analyzer; neither these nor archive hashes establish the independent truth of the exported observations.
Original code/artwork/worksheets: MIT. Original exported observations and calculations: CC0-1.0; third-party source code and marks retain their rights. AI-assisted research, code and drafting. No independent external review or physical miner test claimed. Corrections: https://litecoin.watch/corrections/ . Frozen findings do not update when live tool data changes.