Why does a network interface need to have a routing table entry configured to answer ARP requests?

I'm using a fresh minimal Ubuntu server 24.04.1 LTS install. I run these commands as root to set up networking and do some experiments:

apt install -y netcat-traditional tcpdump inetutils-ping

ip netns add ns1
ip netns add ns2

ip link add my_veth1 type veth peer name my_veth2

ip link set my_veth1 up netns ns1
ip link set my_veth2 up netns ns2

ip -n ns1 address add 1.2.3.4 dev my_veth1
ip -n ns1 route add 2.3.4.0/24 dev my_veth1

ip -n ns2 address add 2.3.4.5 dev my_veth2

Then I run these commands in different terminals:

# Terminal 1
ip netns exec ns1 tcpdump -l -i my_veth1

# Terminal 2
ip netns exec ns2 tcpdump -l -i my_veth2

# Terminal 3
ip netns exec ns1 ping 2.3.4.5

I get the same output in terminals 1 and 2:

02:40:27.511438 ARP, Request who-has 2.3.4.5 tell 1.2.3.4, length 28
02:40:27.511438 ARP, Request who-has 2.3.4.5 tell 1.2.3.4, length 28
02:40:27.511438 ARP, Request who-has 2.3.4.5 tell 1.2.3.4, length 28
...

veth2 has the IP address 2.3.4.5 and is receiving the ARP request. Why doesn't it send an answer? It only answers when I configure a routing table entry:

ip -n ns2 route add 1.2.3.0/24 dev my_veth2

But it shouldn't be necessary since the MAC address of the network interface that veth2 should respond to is already encoded in the request it is responding to.