Rebuild: a file's identity is computed once per batch, not per article

The file key and release identity (two SHA-1 hashes and the release rules)
were ~25% of the edge worker's time. edge:bench: boneless 100k → 137k
headers/s on one core, ath (2013) 70k → 143k.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

lib/headers/segment.ts

@@ -51,22 +51,39 @@ function common(header: OverviewHeader, groupName: string, articleNumber?: bigin
   }
 }
 
+/**
+ * A file's identity, the same for all of its articles: worth computing once
+ * per file (two SHA-1 hashes and the release rules were ~25% of an edge
+ * worker's time; LOG.md 2026-10-09).
+ */
+export type IdentityCache = Map<string, { fileKey: string; release: ReleaseIdentity | null }>
+
 /** Builds a segment from the subject line; null when the subject is obfuscated. */
 export function segmentFromHeader(
   header: OverviewHeader,
   groupName: string,
-  articleNumber?: bigint | number
+  articleNumber?: bigint | number,
+  cache?: IdentityCache
 ): IngestSegment | null {
   const parsed = header.subject ? parseSubject(header.subject) : null
   if (!parsed || !header['message-id']) return null
   const base = common(header, groupName, articleNumber)
+  const id = `${base.poster}\n${parsed.base}\n${parsed.totalSegments}`
+  let identity = cache?.get(id)
+  if (!identity) {
+    identity = {
+      fileKey: fileKey(base.poster, parsed.base, parsed.totalSegments),
+      release: releaseIdentity({ poster: base.poster, subject: parsed.base, name: parsed.name, weak: false }),
+    }
+    cache?.set(id, identity)
+  }
   return {
     ...base,
-    fileKey: fileKey(base.poster, parsed.base, parsed.totalSegments),
+    fileKey: identity.fileKey,
     subject: parsed.base,
     name: parsed.name,
     totalParts: parsed.totalSegments,
     number: parsed.segment,
-    release: releaseIdentity({ poster: base.poster, subject: parsed.base, name: parsed.name, weak: false }),
+    release: identity.release,
   }
 }