Browser failed to connect within 60s. testem.js not loaded?

[Adharsh-Muraleedharan]

while running ember tests via by splitting and running in parallel using ember exam. Most of my tests fail with "Browser failed to connect within 60s. testem.js not loaded"

[Adharsh-Muraleedharan]

we have around 3000 unit test cases which take up an execution time of around 30 mins. To reduce the time we are trying to run tests in a parallel by splitting the cases using ember exam , testem.js and running them in parallel.

File: package.json:

"testInCI": "COVERAGE=true ember exam --split=70 --parallel 'inCiMode'

File: testem.js:

'parallel': process.argv.includes('inCiMode') ? -1 : 5,
'launch_in_ci': [
'Chrome'
],
'browser_start_timeout': 60,
'browser_disconnect_timeout': 120,
'browser_args': {
'Chrome': [
'--headless',
'--disable-dev-shm-usage',
'--no-gpu',
'--disable-software-rasterizer',
'--disable-web-security',
'--incognito',
'--disable-gpu',
'--remote-debugging-port=9222',
'--no-sandbox',
'--window-size=1024,768'
]
}

here we are trying to split the cases in to 70 partitions and trying to run them on parallel. We are directly running the tests on a AWS launched EC2 instance rather than on a Docker container.
while running the tests most of our cases fail with "Browser failed to connect within 60s. testem.js not loaded" error.
Our execution runs on a 36 core machine(c4.8x) and we are not seeing the load average on the machine increase to more than 25.
we tried reducing the split on the cases but it increases the time of our execution tremendously.

From searching through internet we couldn't find a clear solution. Would like to get your suggestions on the below

1. Would switching to a dockerized execution environment solve this issue ? If so how efficiently utilise the containers without this browser failed problem.
2. Suggest on how to avoid this issue with minimal execution time

Thanks for your input in advance!

[raido]

Possible related? #1021