TowerSync Vote Batching: A Historical Source Study and Audit of Public Performance Evidence

This historical study examines an Agave patch that delays vote submission until a count of votable banks reaches a configurable batch size, with exceptions for fork decisions and impending leadership, plus adaptive selection from active stake. Five checks over source-extracted rules passed. They are unit-level checks, not a full validator build or consensus validation. A read-only public RPC audit independently reproduced one cited epoch 978 inflation reward, but all three signatures presented as TowerSync examples returned transactions containing no vote-program instruction. The cited 31-slot root gap is also not a unique batching fingerprint under the upstream Tower rule. Executable configuration defaults to N=1 when omitted, contrary to an adaptive-mode commit message. The public pull request closed without merging. The evidence supports an inspectable submission-policy implementation and identifies gaps in its performance record; it does not establish realized fee savings, net profit, lossless credits, consensus safety or financial inclusion.Jarett Dunn. Technical report, version 1.0. Preprint, not peer reviewed. Codex assisted evidence recovery, execution and writing. Empowerment, financial inclusion and evidence-based accountability guide evaluation; social outcomes are not measured. The manuscript and new analysis are CC BY 4.0; archived software retains its original licenses.