Productivity is always inherently better than the other modules, and that's because the productivity across the entire chain becomes geometric. If you have modules that reduce the power consumption of each building by 40%, your entire factory uses 40% less power. If you have modules that give a 40% speed boost to every building, your factory gets a 40% speed boost. If you have modules that give a 40% productivity boost to every building, your factory gets a 2000% productivity boost. Even if that productivity boost comes with a 6× energy cost, that's
still a 3× boost to items produced per unit energy.
Anyone who comes to understand the geometric nature of productivity then knows there's no point using anything other than productivity modules in machines and speed modules in beacons. A geometric function always beats out a linear function at some point, so making productivity cost more in speed, power, and pollution doesn't change the tradeoff, it just pushes the breakeven point a little further. It's worth pointing out that in vanilla, mixing 3 beacons with speed modules for each assembler with productivities is lower power consumption overall than just the productivity modules (even including the cost of the beacons!).
You're still caught up on this idea that the productivity gain is only a gain, and you're forgetting the drawback--the rate of production. You don't just reach a resource patch and find the ore neatly bundled up in chests, ready to go. It takes time and energy to mine it, more than anything time, and in that time you could easily set up more mining outposts. Resources are, in a way, unlimited. Increasing the amount you get from the patch does not result in suddenly free wealth out of nowhere, rather it simply makes the patch last longer. If you're reducing the rate of output with productivity modules, then you need more infrastructure, not less, to get the same final rate of outputs. In my suggestion, the speed module is the only module which actually increases the rate at which a machine outputs products, and it simultaneously increases the rate the machine consumes its inputs. To determine if the speed module is worth using, you should primarily look at its material cost versus the cost of putting down more machines. But I feel I've made it abundantly clear that the old problem (speed+productivity=faster than only speed) is very specifically not the case in my suggestion. And since your entire argument can only work if it is the case, I must surmise that you are forgetting what I have stated multiple times in this thread already, or that you haven't read my posts which you are trying to disagree with.
The most boring part of Factorio, IMHO, is having to set up new outposts for mining. You have to drive to an appropriate mine, set up the rail tracks to it (woe be he doing this without FARL), drop mining blueprints on the patch, drop the station blueprint, hook all the belts up (if belt-based mining), hook all the track up, set up the station, and then go back to the base. Placing down new solar grids is also boring, but at least I can open up map view, scroll to the edge of my base, and drop down several instances of the solar field and let bots get to it in their own time, it doesn't require that much thought on my own time--or I could have fun trying to design large-scale nuclear power plants (UPS be damned). The logistical challenge of Factorio is the fun part, so saying "resources are unlimited, so why save resources when you have to deal with logistical consequences of over production" isn't going to convince people that productivity isn't powerful.
If you think that it's a problem that people only care to use productivity, you need to find some way to break the productivity-is-geometric, everybody-else-is-linear aspect of modules. Multiplicative effects of multiple modules rather than linear is one way, but the optimal combination is still findable by brute force in very short order. Gating higher-tier module production on something that's truly rare (along the lines of alien artifacts, although those had the annoying problem that they were rare in the midgame when you needed them desperately for research and nothing else and way too common in the endgame, when you couldn't use them fast enough because the only thing you needed was modules) and not amenable to productivity bonuses is another idea (since the modules are too rare to put in everything, you have to strategize about what you actually want to use them for), and probably the one most worth investigating.
It's possible to give people more reasons to use modules much earlier on the game, but that doesn't really solve the geometric/linear discrepancy. Nerfing power production is just asking for another bots/belts brouhaha, especially because it's ultimately harming everybody just to inflict pain on the group of people most likely to shrug it off. And again, it's not touching that discrepancy, just trying to make one factor more painful. As it stands, modules are an endgame-only resource, where things like the cost of production and the difficulty of building power can easily be shrugged off (or factored as just the "cost" of building a megabase). Rebalancing them requires considering their effects at endgame level and the effects in the midgame are damn near inconsequential. As bots/belts shows, you get a lot more positive feedback from the community if instead of asking "how do we punish people for playing the game in the wrong way?" you ask "how can we open up new opportunities for playing the game differently?"
So here's a random idea: what if productivity modules could only be used in buildings that required some form of active cooling using heat pipes? I don't know if mods can do this right now, but it does give more reason to use the heat system which I think is underused in general. It also adds extra constraints in building high-productivity factories which I suspect would be enjoyable to try to work out rather than annoying to deal with, but I obviously haven't play-tested.
