经过几个小时的调试，我找到了罪魁祸首。
ISR页面使用回退策略，如{fallback: true}，因此，它尝试请求HTML页面，如果不可用，它将再次访问相同的路由来获取数据。
为了避免这个问题，我发现NextJS使用头来区分请求的类型。下面是我的解决方案：

import { linksMiddleware } from 'middleware/links';
import { NextFetchEvent, NextRequest, NextResponse } from 'next/server';

export async function middleware(
  request: NextRequest,
  fetchEvent: NextFetchEvent
) {
  // These methods are used when accessing locale routes, so we discard them
  if (request.method === 'OPTIONS' || request.method === 'HEAD') {
    return NextResponse.next();
  }

  //Capture URL locale (if any) and id. For example: /ar/random-id or /random-id
  const pathParams = request.nextUrl.pathname.split('/');

  if (pathParams.length === 2 && isDataRequest(request)) {
    // This means that we are accessing the entity homepage
    return linksMiddleware(request, fetchEvent, 'entity');
  }

  return NextResponse.next();
}

export const config = {
  matcher: [
    // This is to prevent the middleware from running on the API routes and during development
    '/((?!api|_next/static|_next/image|favicon.ico|favicon.svg).*)',
  ],
};

/**
 * Check if the request is a data request for cases of fallback pages
 * This header is absent in the initial request
 */
const isDataRequest = (request: NextRequest): boolean =>
  request.headers.get('x-nextjs-data') === '1' ||
  // The header is absent in development mode since there's no fallback logic (SSR-like)
  process.env.NODE_ENV === 'development';

此外，我还添加了一个HTTP方法检查，用于区域设置路由的情况，不包括OPTIONS和HEAD。