pfSense

in rules.debug:
pass in log quick on $Untrust inet6 proto tcp from $RDEGW01 to <negate_networks> port 25 tracker 10000001 flags S/SA keep state label "NEGATE_ROUTE: Negate policy routing for destination"
pass in log quick on $Untrust $GWVPS_TUNNELV6 inet6 proto tcp from $RDEGW01 to any port 25 tracker 1627229557 flags S/SA keep state label "USER_RULE: SMTP-Pv6"
pass in log quick on $Untrust inet proto tcp from $RDEGW01 to <negate_networks> port 25 tracker 10000002 flags S/SA keep state label "NEGATE_ROUTE: Negate policy routing for destination"
pass in log quick on $Untrust $GWVPS_TUNNELV4 inet proto tcp from $RDEGW01 to any port 25 tracker 1627228080 flags S/SA keep state label "USER_RULE: SMTP-Pv4"

The gateway is up, the alias resolves to the right address. I also tried the ipv6 address of the host explicitely and any, but the result is the same, IPv4 traffic is routed correctly, IPv6 is not.

The VPS interface is GRE tunnel that uses the INTERNET interface.

IPv6 Rule:
<rule>
<id></id>
<tracker>1627229557</tracker>
<type>pass</type>
<interface>Untrust</interface>
<ipprotocol>inet6</ipprotocol>
<tag></tag>
<tagged></tagged>
<max></max>
<max-src-nodes></max-src-nodes>
<max-src-conn></max-src-conn>
<max-src-states></max-src-states>
<statetimeout></statetimeout>
<statetype><![CDATA[keep state]]></statetype>
<os></os>
<protocol>tcp</protocol>
<source>
<address>RDEGW01</address>
</source>
<destination>
<any></any>
<port>25</port>
</destination>
<log></log>
<descr><![CDATA[SMTP-IPv6]]></descr>
<gateway>VPS_TUNNELV6</gateway>
<created>
<time>1627229557</time>
<username><![CDATA[]]></username>
</created>
<updated>
<time>1627233972</time>
<username><![CDATA[]]></username>
</updated>
</rule>

IPv4 Rule:
<rule>
<id></id>
<tracker>1627228080</tracker>
<type>pass</type>
<interface>Untrust</interface>
<ipprotocol>inet</ipprotocol>
<tag></tag>
<tagged></tagged>
<max></max>
<max-src-nodes></max-src-nodes>
<max-src-conn></max-src-conn>
<max-src-states></max-src-states>
<statetimeout></statetimeout>
<statetype><![CDATA[keep state]]></statetype>
<os></os>
<protocol>tcp</protocol>
<source>
<address>RDEGW01</address>
</source>
<destination>
<any></any>
<port>25</port>
</destination>
<log></log>
<descr><![CDATA[SMTP-IPv4]]></descr>
<gateway>VPS_TUNNELV4</gateway>
<created>
<time>1627228080</time>
<username><![CDATA[]></username>
</created>
<updated>
<time>1627230159</time>
<username><![]></username>
</updated>
</rule>

table <negate_networks> { 10.0.23.0/24 ::/0 }
If I remove ::/0 it works. Where is this table coming from?

Only thning i can find related in my config is the mobile client ip pools, but that has a perfectly fine IPv6 prefix:
<ipsec>
<client>
<enable></enable>
<radiusaccounting>enabled</radiusaccounting>
<user_source>...</user_source>
<group_source>enabled</group_source>
<auth_groups>Domain Admins</auth_groups>
<pool_address>10.0.23.0</pool_address>
<pool_netbits>24</pool_netbits>
<pool_address_v6>:0300::</pool_address_v6>
<pool_netbits_v6>64</pool_netbits_v6>
<net_list></net_list>
<save_passwd></save_passwd>
<dns_domain>wizdom.nu</dns_domain>
<dns_split>wizdom.nu</dns_split>
<dns_server1>::aaaa</dns_server1>
<dns_server2>::bbbb</dns_server2>
<dns_server3>.30</dns_server3>
<dns_server4>.31</dns_server4>
</client>

More important question, where does pfsense get the idea that it should make untransparent unlogged routing decisions?

https://github.com/pfsense/pfsense/pull/4534
It is not ok to require end users who are not usually software developers to fix blatently obvious bugs in your software.

I gave you the exact information to locate this within 15 minutes, unless either:
a. You do not know the inner workings of your own software.
b. You are not willing to admit to such obvious bugs.