Hack: tweak deserialisation preference to make integration tests pass

lib/core/src/Cardano/Wallet/Primitive/Types/Tx.hs

@@ -607,8 +607,24 @@ cardanoTxFromBytes
     -> ByteString -- ^ Serialised transaction
     -> Either DecoderError (InAnyCardanoEra Cardano.Tx)
 cardanoTxFromBytes maxEra bs = asum $ map snd $ filter (withinEra maxEra . fst)
-    [ deserialise BabbageEra  Cardano.AsBabbageEra
-    , deserialise AlonzoEra  Cardano.AsAlonzoEra
+      -- NOTE: Attempting to deserialise in Alonzo before Babbage is intentional.
+      --
+      -- It seems Alonzo transactions can be deserialized as Babbage
+      -- transactions. This causes new-tx workflow integration tests to fail.
+      -- The txs are interpreted as Babbage txs right before submitting them to
+      -- the Alonzo-era ledger.
+      --
+      -- This may also cause problems with our store of pending txs.
+      --
+      -- TODO: Can we come up with a better solution?
+      -- - Using Alonzo txs once we're in Babbage doesn't sound good?
+      -- - Ideally we should set @maxEra == currentNodeEra@ argument.
+      --    - But we can't access currentNodeEra from PersistField instances
+      --    - It might be good to ensure txs in the DB also store which era they
+      --      correspond to.
+    [ deserialise AlonzoEra  Cardano.AsAlonzoEra
+    , deserialise BabbageEra  Cardano.AsBabbageEra
+
     , deserialise MaryEra    Cardano.AsMaryEra
     , deserialise AllegraEra Cardano.AsAllegraEra
     , deserialise ShelleyEra Cardano.AsShelleyEra