For example, the "Ambrosia" effect, far from being a nice and flavorful mass healing spell, can become:
- a brutal turn-to-wood attack (instantly lethal?) with "Imbue" execution (and optionally the "Infect" enhancement to lignify multiple people)
- a dangerous bombardment of bushes with Falling and Multiple
- a straightforward (but even sillier) attack with Projectile and Seek
- completely useless with the Instant execution (the bush disappears before anyone can pick the healing fruit).
I think three approaches could give more reasonable results:
- Expand your combinations of spell attributes into gigantic tables and think hard of each entry, which could be forbidden (with or without sarcastic remarks for attempting something stupid), a specific ad-hoc spell (possibly adjusted to make sense or nerfed or enhanced relative to the obvious expectations), or sometimes a spell built from reusable components. More or less the Nethack approach: "the dev team thinks of everything".
- Design a set of sensible and balanced spells, then assign rune combinations to them in the role of magic formulas or component lists. A meaningful rune language could let the players guess spells before they are officially taught to them and collect runes with an idea of their purpose, which should be fun (but not a spell customization system).
- Design spells with customization facets that make sense, without allowing arbitrary runes. For example the Ambrosia bush can be altered to become a bigger bush, a hedge (wall) of any length, multiple bushes or hedges; it can have any duration (including permanent), but it cannot be instantaneous and short durations suck; it can bear more, or more potent, healing fruit; it can be generalized to a bush of magical fruit that do something other than healing; and its power, which should be the starting point to determine its cost, is measured by the aggregate amount of healing it provides, which translates to counting fruit.