Bug with the mod package And GHC 9.4.4 on Linux

Main.hs

{-# LANGUAGE DataKinds #-}
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
{-# LANGUAGE TypeApplications #-}
{-# OPTIONS_GHC -fno-warn-orphans #-}

module Main (main) where

import Data.Mod (Mod)
import GHC.TypeNats (KnownNat)
import Numeric.Natural (Natural)
import Test.QuickCheck (Arbitrary, arbitrary)
import Test.QuickCheck.Instances ()

newtype Fq = Fq (Mod 18446744073709551616 {- 2^64 -}) deriving (Num)

instance KnownNat n => Arbitrary (Mod n) where
    arbitrary = fromIntegral <$> arbitrary @Natural

instance Arbitrary Fq where
    arbitrary = Fq <$> arbitrary

main :: IO ()
main = let Fq x = -1 in print x

mod-bug.caba

cabal-version: 3.0
name: mod-bug
version: 0.1.0.0
synopsis: demonstrate bug when using mod package
executable bug
  main-is: Main.hs
  hs-source-dirs: .
  default-language: Haskell2010
  build-depends:
    , base ==4.17.0.0
    , mod ==0.2.0.1
    , QuickCheck ==2.14.2
    , quickcheck-instances ==0.3.28

With optimization enabled (-O1 or -O2) this program prints

0

which is wrong. Without optimization or with ghci it prints the correct result:

18446744073709551615

The behavior is actually somewhat non-deterministic, at least on MacOS. Some builds give the correct result and some builds give the wrong result. On Linux the wrong result is produced more consistently.

I haven't yet been able to find a minimal example that doesn't include the QuickCheck instances. In a larger production system the bug was observed on a code base without these instances.