HEAD regression: lookupIdSubst panic with GHCi

(Originally observed in a head.hackage issue: head.hackage#88 (comment 513332))

GHC HEAD (at commit eb1a6ab1) panics when loading the following code into GHCi:

-- A.hs
{-# LANGUAGE CPP #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE DefaultSignatures #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE UndecidableInstances #-}
module A where

import Data.Kind

type Sing :: k -> Type
#if __GLASGOW_HASKELL__ >= 909
type family Sing @k
#else
type family Sing
#endif

type SBool :: Bool -> Type
data SBool b where
  SFalse :: SBool False
  STrue  :: SBool True
type instance Sing @Bool = SBool

type DefaultEq :: k -> k -> Bool
type family DefaultEq a b where
  DefaultEq a a = 'True
  DefaultEq a b = 'False

type Not :: Bool -> Bool
type family Not b where
  Not False = True
  Not True  = False

sNot :: forall (b :: Bool). Sing b -> Sing (Not b)
sNot SFalse = STrue
sNot STrue  = SFalse

class PEq a where
  type (x :: a) == (y :: a) :: Bool

  type (x :: a) /= (y :: a) :: Bool
  type x /= y = Not (x == y)

class SEq a where
  (%==) :: forall (x :: a) (y :: a).
           Sing x -> Sing y -> Sing (x == y)

  (%/=) :: forall (x :: a) (y :: a).
           Sing x -> Sing y -> Sing (x /= y)
  default (%/=) :: forall (x :: a) (y :: a).
                   ((x /= y) ~ Not (x == y)) =>
                   Sing x -> Sing y -> Sing (x /= y)
  (%/=) sx sy = sNot (sx %== sy)

-- B.hs
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE UndecidableInstances #-}
module B where

import GHC.Exts (TYPE)
import Type.Reflection

import A

type instance Sing @(TYPE rep) = TypeRep

instance PEq (TYPE rep) where
  type x == y = DefaultEq x y
instance SEq (TYPE rep) where
  tra %== trb =
    case eqTypeRep tra trb of
      Just HRefl -> STrue
      Nothing -> undefined

$ ghc-9.9.20230716 --interactive B.hs
GHCi, version 9.9.20230716: https://www.haskell.org/ghc/  :? for help
Loaded GHCi configuration from /home/rscott/.ghci
[1 of 2] Compiling A                ( A.hs, interpreted )
[2 of 2] Compiling B                ( B.hs, interpreted )

<no location info>: error:
    panic! (the 'impossible' happened)
  GHC version 9.9.20230716:
        lookupIdSubst
  eta_i1Ia
  InScope {rep_a1uz $c%==_a1uB $c%/=_a1GR x_a1GW y_a1GX a_i1HX
           $dSEq_i1HY x_i1HZ y_i1I0 $d~_i1I1 eta_i1I2 eta1_i1I3 co_i1I4
           $fSEqTYPE $fPEqTYPE}
  Call stack:
      CallStack (from HasCallStack):
        callStackDoc, called at compiler/GHC/Utils/Panic.hs:191:37 in ghc-9.9-inplace:GHC.Utils.Panic
        pprPanic, called at compiler/GHC/Core/Subst.hs:197:17 in ghc-9.9-inplace:GHC.Core.Subst
  CallStack (from HasCallStack):
    panic, called at compiler/GHC/Utils/Error.hs:503:29 in ghc-9.9-inplace:GHC.Utils.Error


Please report this as a GHC bug:  https://www.haskell.org/ghc/reportabug

Failed, one module loaded.

GHC 9.6 and earlier do not panic, on the other hand. I was only able to trigger the panic with GHCi, and not with an ordinary invocation of GHC.