[ OCP 4.12] IPXE connection timed out

New deployment of BM IPI using provisioning network with IPV6 is showing:

http://XXXX:XXXX:XXXX:XXXX::X:6180/images/ironic-python-agernt.kernel....
connection timed out (http://ipxe.org/4c0a6092)" error

Openshift 4.12.32
Also seen in Openshift 4.14.0-rc.5 when adding new nodes

Very frequent

1. Deploy cluster using BM with provided config
2.
3.

Consistent failures depending of the version of OCP used to deploy

No error, successful deployment

Things checked while the bootstrap host is active and the installation information is still valid (and failing):
- tried downloading the "ironic-python-agent.kernel" file from different places (bootstrap, bastion hosts, another provisioned host) and in all cases it worked:
[core@control-1-ru2 ~]$ curl -6 -v -o ironic-python-agent.kernel http://[XXXX:XXXX:XXXX:XXXX::X]:80/images/ironic-python-agent.kernel
\*   Trying XXXX:XXXX:XXXX:XXXX::X...
\* TCP_NODELAY set
% Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                               Dload  Upload   Total   Spent    Left  Speed
0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0* Connected to XXXX:XXXX:XXXX:XXXX::X (xxxx:xxxx:xxxx:xxxx::x) port 80   #0)
> GET /images/ironic-python-agent.kernel HTTP/1.1
> Host: [xxxx:xxxx:xxxx:xxxx::x]
> User-Agent: curl/7.61.1
> Accept: */*
>
< HTTP/1.1 200 OK
< Date: Fri, 27 Oct 2023 08:28:09 GMT
< Server: Apache
< Last-Modified: Thu, 26 Oct 2023 08:42:16 GMT
< ETag: "a29d70-6089a8c91c494"
< Accept-Ranges: bytes
< Content-Length: 10657136
<
{ [14084 bytes data]
100 10.1M  100 10.1M    0     0   597M      0 --:--:-- --:--:-- --:--:--  597M
\* Connection #0 to host xxxx:xxxx:xxxx:xxxx::x left intact

This verifies some of the components like the network setup and the httpd service running on ironic pods.

- Also gathered listing of the contents of the ironic pod running in podman, specially in the shared directory. The contents of /shared/html/inspector.ipxe seems correct compared to a working installation, also all files look in place.

- Logs from the ironic container shows the errors coming from the node being deployed, we also show here the curl log to compare:

xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx - - [27/Oct/2023:08:19:55 +0000] "GET /images/ironic-python-agent.kernel HTTP/1.1" 400 226 "-" "iPXE/1.0.0+ (4bd064de)"
xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx - - [27/Oct/2023:08:19:55 +0000] "GET /images/ironic-python-agent.kernel HTTP/1.1" 400 226 "-" "iPXE/1.0.0+ (4bd064de)"
xxxx:xxxx:xxxx:xxxx::x - - [27/Oct/2023:08:20:23 +0000] "GET /images/ironic-python-agent.kernel HTTP/1.1" 200 10657136 "-" "curl/7.61.1"
cxxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx - - [27/Oct/2023:08:20:23 +0000] "GET /images/ironic-python-agent.kernel HTTP/1.1" 400 226 "-" "iPXE/1.0.0+ (4bd064de)"
xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx - - [27/Oct/2023:08:20:23 +0000] "GET /images/ironic-python-agent.kernel HTTP/1.1" 400 226 "-" "iPXE/1.0.0+ (4bd064de)"

Seems like an issue with iPXE and IPV6