Incremental parse resumes from wrong position after file shrink/rewrite

[httpsVishu]

noticed a stale-state issue around same size transcript rewrites in src/providers/cursor/transcripts.ts file

right now tryCacheHit() treats the file as unchanged whenever sizeBytes matches the cached value, so if cursor rewrites/regenerates a transcript but the resulting file happens to have the same byte size, the parser skips reparsing entirely and just reuses the old cached parse state

const contentChanged = fileSizeBytes !== cached.sizeBytes;

if (!contentChanged) {
  return {
    result: {
      agents: resolveAgentsFromState(cached.state, sourcePath, cached.fileUpdatedAt, now),
      success: true,
      warnings: [],
    },
    updatedCache: { ...cached, mtimeMs: fileUpdatedAt },
  };
}

there is also a related assumption in resolveParseStrategy() where incremental parsing resumes from cached.lineCount whenever size/line count are at least the cached values which only seems safe for strictly append-only transcripts

in practice, if a session file gets rewritten/reset with different content but identical byte length, stale agents/task state from the previous session can persist until the file size changes again, this looked especially risky for restarted/re-generated cursor conversations

was append-only behaviour guaranteed here or could rewritten transcripts bypass cache invalidation?