[Affected versions]:
Nightly 79.0a1 (2020-06-22)
Beta 78.0b9 (64-bit)
Release 77.0.1

[Affected platforms]:
MacOs 10.14

[Preconditions]
Have devtools.responsive.browserUI.enabled = true

Steps :

1- Launch the Firefox browser
2- Reach youtube
3- Hit Cmd+option+M in order to start Responsive Design Mode, have both "reload when touch simulation is toggled" and "reload when user agent is changed" CHECKED, and Enable Touch Simulation.
4- resize viewport from bottom right corner
5- notice that it resizes properly and video keeps working
6- click on any video
7- now choose any device from the list, size adjusts accordingly.
8- notice that youtube switched to mobile version
9- resize again, device is not selected anymore now it's " responsive" by default

Expected Results :
videos should play normally

Actual Results :
nothing happens when clicking play, user needs to switch back to desktop version from the three dots menu.

Okay, I can reproduce a version of this. My Steps to Reproduce:

1. Navigate to http://m.youtube.com/?persist_app=1&app=m which will force your browser to display the mobile version of YouTube.
2. Attempt to play one of the videos. I get a download link to an rtsp:// stream.
3. Open RDM and select a mobile device. I chose iPad.
4. Play a video; everything works.
5. Change the viewport size either by dragging the corner or by manually editing the viewport size. The device changes to "Responsive". If the video was playing, it will stop.
6. Try playing the video again, and you get the rtsp:// stream.

When you're done, and you want to switch back to desktop YouTube, visit http://www.youtube.com/?persist_app=1&app=desktop.

This will happen even if the setting "Reload when the user agent is changed" is toggled off. If you copy the iPad user agent string and paste it back in after changing the viewport, the bug will still occur. In other words, there's some detection going on that's based on something other than user agent string. It's also not strictly dependent on viewport size -- in Chrome Device Mode this operation can be done smoothly without stopping the video and without getting kicked back to the rtsp:// stream.

Micah, I'm not sure what else we might be exposing that is being detected by the YouTube site. Do you have any ideas?

(In reply to Brad Werth [:bradwerth] from comment #1)

Micah, I'm not sure what else we might be exposing that is being detected by the YouTube site. Do you have any ideas?

Thanks for investigating this Brad! I'm not aware of anything in RDM that we are exposing that would influence the Youtube site to detect this bug. My first thought is the change in user agent and/or viewport size, which you've noted isn't necessarily what's causing it. I'll take a look into more.