Type variables from instance declarations reinterpreted as existential type variables in associated data family instances

Summary

A simple program results in ghc panic, where instead it should be a compilation error.

[1 of 1] Compiling Main             ( ghc-impossible.hs, ghc-impossible.o )
ghc: panic! (the 'impossible' happened)
  (GHC version 8.8.3 for x86_64-unknown-linux):
	mkTyConKindRepBinds.go(tyvar)
  s_a10z[sk:1]
  Call stack:
      CallStack (from HasCallStack):
        callStackDoc, called at compiler/utils/Outputable.hs:1159:37 in ghc:Outputable
        pprPanic, called at compiler/typecheck/TcTypeable.hs:536:9 in ghc:TcTypeable

Steps to reproduce

Here is a minimal example that triggers the bug

{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE TypeFamilies #-}
import Control.Monad.ST

class Monad m => MonadRandom g m where
  data GenM g :: (* -> *) -> *

type family StateToken (m :: * -> *) :: *
type instance StateToken IO = RealWorld
type instance StateToken (ST s) = s

newtype MutGen g = MutGen g

data MutVar s g

instance (s ~ StateToken m, Monad m) => MonadRandom (MutGen g) m where
  newtype GenM (MutGen g) m = MutGenM (MutVar s g)
-- Correct version below compiles:
--  newtype GenM (MutGen g) m = MutGenM (MutVar (StateToken m) g)

Expected behavior

What do you expect the reproducer described above to do?

Environment

• GHC version used: 8.8.2 and 8.8.3

Optional:

• Operating System:
• System Architecture: