Ask HN: Components to build CrossBrowserTesting/Litmus Clone?

I have a need to do screenshots on actual devices, not only for landing pages (CrossBrowserTesting) but for email clients (Litmus).

CrossBrowserTesting - $1,500 per month.

EmailOnAcid - $1.60 per screenshot.

These are really unsustainable to support a service that I am developing.

So I am looking at the possibility of building my own infrastructure/software to support my own screenshot service.

I just want to get this out of the way, my focus is not on ChromeDP and I don't wish to use it in terms of emulation via user agents. What I want to do is focus on real hardware, os and browser/email clients.

Now with that in mind, what do I need?

1) Real devices? Ios, Android, etc.

2) Server cluster? Array of VMs (Win/Mac/Lin)

3) Browsers - (C, IE/ED, FF, OP, SF)

4) Email Clients - Respective to their OS

5) Email Accounts (Hotmail/Outlook, Gmail, etc)

If there are any other components needed let me know.

But here is where I fall short. What provides the orchestration here? Lets say I have this cluster up and running, connected to 50 mobile devices. Also multiples of browsers running on each the OS's, with browsers and email clients running in foreground/background...

What is the mechanism that allows me to submit either a URL or waits for a specific email and then renders all the screenshots and returns them to a user?

Would love to hear if there are any opensource projects doing something similar here.

If you have actual experience in building something like this from the ground up. I would very much like to hear from you.

Finally, and I suppose this is the most important one. What are the gotchas in maintaining this type of service? Where are the pain-points or challenges to getting this up and running?

For example. Previewmyemail was doing unlimited screenshots for a range of email clients for $248 per month and then has gone offline. They stopped updating their Twitter in May and their Facebook no longer exists. I wonder why?