Nice kata but unfortunately some Elixir test cases are broken, for example:

Testing input: "38691?+-198280=18863?"
1) test Fixed tests More tests (TestSuite)
     test/solution_test.exs:51
     Assertion with == failed
     code:  assert Kata.solve_runes(s) == sol
     left:  0
     right: 8
     stacktrace:
       (elixir 1.11.4) lib/enum.ex:798: Enum."-each/2-lists^foreach/1-0-"/2
       test/solution_test.exs:64: (test)

Proper solution is 4 in this case, which also works on my local machine:

assert(4 = CodeWars.Numbers.solve_runes("38691?+-198280=18863?"))

I agree, there is no good balance in performance constraints between languages.

47848+-?39008=-5911?0
47,848+-(100,000x + 39,008) = -(591,100 + 10x)
47,848 - 100,000x - 39,008 = -591,100 - 10x
-99,990x = -599,940
x = 6