fix(python): de-flake test_invoke_script_with_line_flush

[omarqureshi]

Summary

Managed to get this test to fail whilst working on ruby language bindings.

test_invoke_script_with_line_flush asserted exact rounded inter-line delays (== 0 / == 1), which is brittle on loaded machines: startup jitter or a slow 1s sleep can round a measured delay to an unexpected value and fail the test spuriously.

This splits the assertion into the two properties the test actually exists to verify:

• the output lines match exactly (unchanged in strictness)
• each line printed after a delay(1000) arrived >= 1s after the previous line

The streaming/no-buffering guarantee is preserved: if stdout were block-buffered, all lines would arrive in one burst at process exit and every inter-line delay from index 3 onward would be ~0, failing the new assertion.

The script that this test counts relies on reads as follows:

  console.info('Hello World!');                          // line 1
  ...
  console.info(`  arguments: ${args.join(', ')}`);       // line 2
  if (args.includes('delay')) {
    for (let i = 1; i <= 5; i++) {
      console.log(`sleeping 1s ${i}`);                   // line 3 (i=1) — printed BEFORE the sleep
      await delay(1000);                                 // then sleeps
    }
  }

as such the expected output is:

idx	line	gap	why
0	Hello World!	0	by construction — last_dt is initialized on the first line
1	arguments: arg1, delay	~0	printed in the same burst
2	sleeping 1s 1	~0	printed before the first await delay(1000)
3	sleeping 1s 2	~1s	arrives after the first sleep
4–6	sleeping 1s 3..5	~1s each	one sleep between each

🤖 Generated with Claude Code

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.