The problems with your approach seems to be the problems with MacGyver:
People don't just instantly invent super technology that works extremely well and immediately solves problems and defeats strong opponents, except rarely and in extremely limited conditions that usually involve one-time surprises more than it's that their invention is a new uber-weapon. Like the makers of MacGyver, you seem to have a tendency to allow any inventive-seeming idea to be invented, designed, produced and deployed in a short time, and to work extremely well, overpowering conventional situations. That's over-representation of how easy, quick and effective invented technologies are.
Also, unless you and your players all agree and enjoy that you are playing a MacGyver game, you shouldn't be handling it like it is.
In the molotov cocktail case, you are thinking this sounds like a super-powerful weapon that has advantages over, say, a trained fighter using a sword, but it's not. As others have pointed out, many medieval games include rules for fire bombs, and they're not that great. Without gasoline, they are just auxiliary devices that can sometimes be useful or give an edge, but are clumsy to use and not that effective. Going into combat holding a lit torch and bottles of oil leaves you exposed and likely to be taken down and end up burning yourself, and lamp oil doesn't burn all that hot and will take a while to hurt someone in armor - even if you hit someone, they can finish the fight possibly without much injury and then roll on the ground to put the fire out.
In the case someone mentioned of a "dwarvish engineer" inventing a gun - read up on how many years (lifetimes) it took collective efforts of Europeans to develop useful and reliable firearms, and how bad the early versions were.
In other cases, these are usually clever tricks which will require not just the idea, but developing the idea, various uncommon materials, craftsmanship, trial and error, certain conditions, skill development, lots of uncertainty, lack of disturbance or discovery by opponents, surprise/panic/foolish reactions by victims, etc.
The GM is the one who gets to decide how effective any improvised course of action is.
So, learn historical and realistic perspective and apply it to your estimates of what it will take to realize an idea, and what the odds of success will be. Many ideas may work once, or almost work but instead waste lots of time and effort but not quite work, etc.
And of course, actual good ideas may be good ideas and deserve rewards and results and be fun... and if they are general purpose, may be copied or countered.
There is also the issue of player knowledge versus character knowledge and ability. Again like MacGyver, I suspect you are allowing the players' enthusiasm to translate into character abilities that aren't on their character sheets. MacGyver is a super-hero specialized in inventing ridiculous devices in a few seconds based on encyclopedic knowledge and super crafting skills. None of that, I suspect, is on your players' character record sheets. These things should be explained and agreed to up-front as part of explaining roleplaying expectations. One form is bringing players' 21st-Century knowledge into medieval fantasy worlds. Just as the player can't have their other-world character talk about USA pop music, they can't have them know about chemistry or anything that their character wouldn't know about from the character's own experience. If a player can't handle this, then they aren't capable of roleplaying from the setting, and need to either be restrained or to play a time/world-travelling character, if everyone agreed that would even be allowed. Now, if they actually want to play a plausible in-world character who is highly inventive and educated in rare-but-in-world knowledge, then they just need to use characters where that is appropriate in the game system, with the trade-offs that involves in the game system. I.e. they need to build their characters with the knowledge and brilliance they are going to want the character to have available, rather than allowing any PC to be brilliant while playing a combat-optimized barbarian with no in-game specialization in such things.
You mentioned that you "can hardly keep my players from buying lamp oil or bottles", which is true, but you can require them to have some in-world, in-character information or Galileo-like invention talent which allows them to get the idea to do something, or for the PC's to invent a new idea that the characters don't have.
You might also want to try having the players try to do some things in real life that they think they know about, but have never done, to give you all some experience in what real-world MacGyverism is like. Try having non-cooks try to cook a soufflé or other semi-difficult recipe and experience trying to get it to turn out right. Try to attach a tablet connected to video conferencing software to a radio-controlled helicopter which lacks a camera, and try using it to guide its flight in first-person. Try having a guy with a water balloon try to defeat someone with a Nerf bow-and-arrow or Nerf sword. Try building a catapult. Etc. See how many hours or days and failures it takes. Imagine having to design and build things with no access to modern tools or shops. Etc.