perf(github): review-thread reads cost 6 GraphQL points instead of 14

packages/source-control-github/src/server/GitHubPullRequestApi.test.ts

@@ -3786,18 +3786,23 @@ layer("GitHubPullRequestApi.layer", (it) => {
     }),
   );
 
-  it.effect("stops at the thread bound and says the conversation was cut short", () =>
+  it.effect("stops at a thousand threads and says the conversation was cut short", () =>
     Effect.gen(function* () {
-      mockedExecute.mockReturnValue(
-        Effect.succeed(
+      // Serves whatever page size the document asks for, so the bound holds in threads.
+      mockedExecute.mockImplementation((call) => {
+        const query = call.kind === "graphql" ? call.query : "";
+        const first = Number(/reviewThreads\(first:(\d+)/.exec(query)?.[1]);
+        const start = Number(call.kind === "graphql" ? (call.variables?.["after"] ?? 0) : 0);
+        const ids = Array.from({ length: first }, (_, index) => `PRRT_${start + index + 1}`);
+        return Effect.succeed(
           output(
             reviewThreadsPage(
-              [{ ...thread("PRRT_1", "c1"), comments: threadComments(["c1"], "Y3Vyc29yOjI", 3) }],
-              "Y3Vyc29yOjE",
+              ids.map((id) => thread(id, `c_${id}`)),
+              String(start + first),
             ),
           ),
-        ),
-      );
+        );
+      });
       const cli = yield* GitHubPullRequestApi.GitHubPullRequestApi;
 
       const conversation = yield* cli.listReviewThreadComments({
@@ -3807,7 +3812,8 @@ layer("GitHubPullRequestApi.layer", (it) => {
         number: 7,
       });
 
-      assert.strictEqual(mockedExecute.mock.calls.length, 10);
+      assert.strictEqual(conversation.reviewThreads.length, 1_000);
+      assert.strictEqual(conversation.reviewThreads.at(-1)?.id, "PRRT_1000");
       assert.isTrue(conversation.truncated);
       assert.isTrue(conversation.reviewThreadsTruncated);
     }),

packages/source-control-github/src/server/GitHubPullRequestApi.ts

@@ -366,6 +366,9 @@ export const NODE_ID_CACHE_CAPACITY = 128;
 
 const REVIEW_THREAD_PAGES = 10;
 
+// `ReviewThreads` reads 25 threads a page; forty pages keep the 1,000-thread ceiling.
+const REVIEW_THREAD_LIST_PAGES = 40;
+
 const CHECK_CONTEXT_PAGES = 10;
 
 export interface GitHubPullRequestListBatch {
@@ -2231,7 +2234,7 @@ export const make = Effect.gen(function* () {
         let cursor = first.nextCursor;
         let page = 1;
         // Later pages carry the threads alone; the PR-wide fields above come from the first.
-        while (cursor !== null && page < REVIEW_THREAD_PAGES) {
+        while (cursor !== null && page < REVIEW_THREAD_LIST_PAGES) {
           const later: Pick<
             GitHubReviewThreadPage,
             "threads" | "nextCursor" | "avatarsByLogin" | "botLogins"

packages/source-control-github/src/server/pullRequestReviews.graphql

@@ -89,7 +89,9 @@ query ReviewThreads($owner: String!, $name: String!, $number: Int!, $requestedRe
   }
   repository(owner: $owner, name: $name) {
     pullRequest(number: $number) {
-      reviewThreads(first: 100) {
+      # GitHub prices a page by the rows it may return, not the rows that exist, so a page of
+      # twenty-five threads costs 6 points even on a pull request with none.
+      reviewThreads(first: 25) {
         ...ReviewThreadPage
       }
       viewerCanUpdate
@@ -204,7 +206,7 @@ query ReviewThreadPages($owner: String!, $name: String!, $number: Int!, $after:
   }
   repository(owner: $owner, name: $name) {
     pullRequest(number: $number) {
-      reviewThreads(first: 100, after: $after) {
+      reviewThreads(first: 25, after: $after) {
         ...ReviewThreadPage
       }
     }