For me, at least, the main purpose of pausing the videos is so I could read the subtitles/text, maybe take notes, and give me a moment to "digest" the content. It had not even crossed my mind to try switching to the main game and back -- I never left the video screen. Perhaps you can do both pause and fast forward for now? Just alert the user that if they leave he video while paused, it will reset.
I am wondering now if you can auto-FF to a specific point in the video (and of course, obtain the current position), which you could save in global. That would make it possible to pause, go to main game, come back, and jump back to the saved position. It would also make it possible in a future release to create multiple "bookmarks" players could jump to while learning how to do things. All this, of course, depends on the capabilities of the video library you are using. (I have no experience with A/V playback libraries--my A/V disabilities are too limiting to bother learning.)
Regarding power switches, perhaps do both? Allow direct manual control, but also allow circuit control. Just like in MineCraft, you can manually open most doors, as well as wire RedStone circuits to control their state. Much if this has to do with my original hopes on what this game was, which I will explain...
Originally, I was looking and hoping for a game similar to MineCraft in some ways, yet different in others. Other influences include Colobot and Widelands, and the Itch.io games Autonauts, BaseCraft, Craftomation 101, Mindustry, Villager, and others. The basic idea is that you start with very little and have to build up your abilities: Manual mining the first resources in which to build more machines to mine/produce more advanced machines. Automation is something of a goal: As you build more advanced machines, some of them are useful to create automated lines and circuitry to control automation. Thus, the progression/goal starts with manual resource gather and basic crafting, to build up to a point it becomes automated and self-sustaining.
The rocket can arrive at regular intervals and start by needing basic items (i.e. "1000x copper")... Harder "modes" could add limits, such as "1000x copper, within 10 days. 3 failures to delivery in time == game over". The "needs" get more advanced as you complete them: Complete "1000x copper", and next time they want "1000x copper and 500x copper plates", and so on. This becomes the "push" for the player to build up more advanced machines and production lines, eventually automating them as well.
This also means that full automation becomes a goal, rather than something the player can/will do from the start. It also becomes a reason and incentive to build a"rocket landing pad", and "rewards" for completed goals: "Build a rocket pad, deliver 1000x copper, and get rewarded with an extra extractor, conduit, and solar panel, delivered by the next rocket".
Of course, keep the current game as a "sandbox" mode where the player can experiment and try things.
I.M.O., these kind of things are what would make the game fun and profitable, for both players and yourself. You could add Patreon support and eventually a premium Steam version. Even if you never intend the game to become a "for pofit" game, you can keep the price low enough just to cover your development costs and time. You can start thinking of future plans; Additional machines, "quests"/goals, scenarios, maybe even one day allowing user made mods and scenarios (take a look at Battle for Wesnoth's history and development).