[Bug]: sandbox job with child process is slow

[zshnb]

Version

v5.58.4

Platform

NodeJS

What happened?

I have a complex asr task, it has three steps, split -> transcribe -> merge. I use flow in transcribe and merge steps, in transcribe task I use sandbox job with childprocess, the code is below

await runPythonScript(
	'toolchain/transcriber.py',
	['--segment_file_path', segment]
)

what I find in log is below

as I highlight in red rectangle, nodejs start exec childprocess time compare python script start transcribe is different for 19 seconds. I know transcribe is a cpu heavy task, is there has some way to speedup it?

How to reproduce.

No response

Relevant log output

Code of Conduct

• I agree to follow this project's Code of Conduct

[zshnb]

more context
this is nodejs execute child process

export async function runChildProcess(cmd: string, args: string[], timeout: number) {
  const fullCommand = `${cmd} ${args.join(' ')}`
  logger.info(`executing ${fullCommand}`)
  const process = spawn(cmd, args, {
    timeout,
  })
  const start = Date.now()
  return new Promise((resolve, reject) => {
    process?.stdout?.on('data', (data: Buffer) => {
      data
        .toString()
        .split('\n')
        .forEach((line: string) => {
          if (line) {
            logger.info(`stdout: ${line.trim()}`)
          }
        })
    })
    process.stderr?.on('data', (data: Buffer) => {
      data
        .toString()
        .split('\n')
        .forEach((line: string) => {
          if (line) {
            logger.error(`stderr: ${line.trim()}`)
          }
        })
    })

    process.on('error', (err) => {
      const error = err as Error
      logger.error(err, `${fullCommand} error: ${error.message}`)
      reject(new Error(error.message))
    })

    process.on('exit', (code) => {
      logger.info(`${fullCommand} exited with code ${code}, cost ${(Date.now() - start) / 1000}s`)
      if (code !== 0) {
        reject(new Error(`${fullCommand} exited with non-zero code`))
      } else {
        resolve(Date.now() - start)
      }
    })
  })
}

logger.info(f'transcribe_segment: {segment_file} with options: {options}'), this log is in the first line of python script