D&D 5E Balancing "consumable" powers vs "per time" powers (Alchemists, Grenadiers etc) - the search for an alternative

[Saeviomagy]

My go-to for running any kind of technologist would be the concept that anything that isn't mass produced and consumer focused requires skill to use and maintain.

Don't want your grenadier handing out grenades to the entire party? Using a grenade isn't as simple as light, throw. You might have to make sure the powder is settled right, the wick is at the right depth, the throw has to make sure it doesn't upset all of that, and of course getting it lit requires you to blow on the wick just the right amount and in the right direction as you apply the flame.

Don't want your grenadier holding 200 grenades in his backpack? Poorly kept gunpowder goes 'off', so each grenade needs topped up every day or you risk premature detonation. Damp powder needs removed. The wax protecting the container from corrosion needs inspected and replaced. The wick needs to be replaced if it soaks up phlostigon etc etc.

Long story short - you can either say "no, only the grenadier can use this stuff, and he can only maintain X points of stuff at a time" or you can let the party bend and break those rules with appropriate consequences. For every item beyond your level limit, you get a 5% misfire chance. For each round after obtaining a grenade from the grenadier, you get a 5% misfire chance. etc.

 

[Fanaelialae]

What you're suggesting is essentially a class designed around x/level or x/campaign abilities. While that certainly can be implemented, it's impossible to precisely balance against an x/rest class because you can potentially have an infinite number of rests inside of a level or campaign. The best you can perhaps do is estimate the number based on the expected number of rests per level or campaign. And then add a bit extra, since having a little more than they strictly need is better than the player having nothing to do because the dice were cold and a few more rests were required than average. Ultimately, it's not a basis for class design that I would use, since for me it would be more work than it's worth, but I think it is feasible if you can get accurate play data on the timeframe you intend to use.