cli.rb

module ReportCli
  class Runner

    include Concurrent::Async
    
    attr_accessor :threads_num, :run_times, :caller

    def initialize
      @conn = Config.connection
    end

    # TODO use TPoolexecutor to limit queue ,
    # http://ruby-concurrency.github.io/concurrent-ruby/Concurrent/ThreadPoolExecutor.html

    def run
      pool = Concurrent::FixedThreadPool.new(self.threads_num)
      
      @instance = self
      
      pool.post{ 
        @instance.run_times.times{ 
          @instance.caller.exec 
        } 

      }
      pool.wait_for_termination
      
    end

    def call
      puts "sending reports"
      self.caller.send_report
    end

  end

end

example.rb

      runner  = PreyReportCli::Runner.new
      runner.threads_num = 5
      runner.run_times = 5
      runner.caller = PreyReportCli::Report.new
      runner.run

report.rb

module ReportCli
  class Report
    
    def initialize
      @conn = Config.connection
    end

    def end_point
      "/api/v2/devices/#{Config.device_key}/reports.json"
    end

    def exec
      payload = {}
      payload[:screenshot] = Faraday::UploadIO.new("./#{shuffled_images}.jpg", 'image/jpeg')
      @conn.post end_point, payload
    end

    def shuffled_images
      ["mat","pic","van"].shuffle.first
    end

  end

end

I'll post a gist of an updated script in a few minutes, but here are a few notes:

• include Concurrent::Async makes the class inherently asynchronous. It will allow its methods to be run in the background on the global thread pool and will ensure that they are synchronized and run in order. It is not necessary in this context (nor are you actually using it).
• pool.wait_for_termination is not used to wait for all jobs to finish, it's used for waiting until the pool shuts down. It will block forever unless you call pool.shutdown or pool.kill, neither of which you are calling.
• pool.wait_for_termination won't work in this context. Calling #shutdown and #kill cause the pool to throw away unprocessed jobs. This is not what you want.
• I would recommend that you do not limit the number of threads in the pool. Our thread pools dynamically grow and shrink as needed. When doing a lot of IO it's generally much more efficient to let the pool itself decide how big it needs to grow.

Also, your Report class may not be thread safe. I don't know what you are storing in Config.connection but that needs to be thread safe for this to work. If it isn't thread safe, your code isn't thread safe. I'm going to assume that it is thread safe.

@michelson Please take a look at this: https://gist.github.com/jdantonio/baf400950605e7e01f36