Dead-lettering from a QQ fails when upgrading from 4.0 to main

[mkuratczyk]

Describe the bug

NOTE: this issue is only applicable to unreleased version of RabbitMQ (upgrading to an alpha version of RabbitMQ 4.1)

When upgrading from 4.0.5 with a QQ that has a DLQ configured and receives messages with TTL set (which are correctly dead-lettered on 4.0.5), when the first node in the cluster is upgraded to main, the QQ crashes:

  initial call: ra_server_proc:init/1
  pid: <0.2357.0>
  registered_name: '%2F_publish_only_ttl'
  exception error: no function clause matching lists:zip("娵",[],fail) (lists.erl, line 678)
    in function  rabbit_fifo_dlx:'-discard/4-fun-2-'/7 (rabbit_fifo_dlx.erl, line 161)
    in call from ra_server_proc:handle_effect/5 (src/ra_server_proc.erl, line 1415)
    in call from lists:foldl_1/3 (lists.erl, line 2151)
    in call from ra_server_proc:handle_effects/5 (src/ra_server_proc.erl, line 1323)
    in call from ra_server_proc:follower/3 (src/ra_server_proc.erl, line 843)
    in call from gen_statem:loop_state_callback/11 (gen_statem.erl, line 3735)

Reproduction steps

cd v4.0.x
make start-cluster
rabbitmqadmin declare policy name=ttl pattern='qq-.*' definition='{"dead-letter-exchange":"","dead-letter-routing-key":"ttl_dlq"}' apply-to=queues
perf-test -u ttl_dlq -qq -C 1 -c 1 # declare the DLQ
perf-test -x 100 -y 0 -qq -c 1 -qp qq-%d -qpf 1 -qpt 100 --message-properties expiration=10000 -qa x-queue-leader-locator=balanced

cd ../main
make restart-cluster

Expected behavior

no crashes

Additional context

No response