Spent some time discovering and unfortunately it's Windows's bug in loopback interface.
I wrote a test demo (blocking mode) to exchange data and watched their packets, found that network stack would drop packets when congestion control algorithm is set to BBR2. It seems the second data exchange was broken.

To avoid further noise on this ticket, i've done as requested and posted to gnupg-devel : https://lists.gnupg.org/pipermail/gnupg-devel/2025-May/035875.html

For the records:

A bug tracker shall never be used for discussion because the audience is not as expected. Only very few people follow a certain bug but several hundreds are following discussion on gnupg-devel@. That is basic hacker knowledge.

I am surprised that you don't want to use the issue tracker for issues.
GnuPG's trust calculations are quite clearly broken, by any metric. There's nothing to discuss here.

Heiko, I told you already in T7106 that it is not a good idea to re-open a ticket. If you really want to discuss stuff, take that to a mailing list.

I see two interesting angles from which to think about this Web of Trust calculation:

> I'm not sure i understand why "the latest" should be preferred.

A bit more experimentation shows the same behavior, even if Alice's tsig of Bill is full, not marginal, and even if all signatures are made in the same second, which is the finest resolution that OpenPGP objects can report.

Interesting analysis, thanks for the sleuthing! I'm not sure i understand why "the latest" should be preferred. For example, in the graph made in this example, which part of the graph is the "latest"? Since the path from Alice to Carol is two hops long at least, it's conceivable that one path (A→Bob→C) has both "the latest" tsig *and* "the earliest" tsig, if the other path (A→Bill→C) happens to have been made between the other two tsigs.

I also spend some time with this and the problem is described by this comment in trustdb.c:

No, it is not a bug and I beg you not to change the status again. Don't start the same trouble here as some of you guys did with the IETF WG!

Err, I don't see why I would "need to test" anything further.

This is just one build of PGP and you would need to test all versions on Windows, macOS and Unix. You also need to test against all versions of GnuPG since 1998 (when we started with interop tests). We won't change this in GnuPG and risk regression. If you have a problem with that go and add a fix to your tool - name it bug compatibility or whatever. And please do not re-open this bug.

In T7106#185462, @werner wrote:

This has been implemented and tested to be compatible with PGP - a looong time ago. iirc this was discussed around 1999 but might be only by private mail between the PGP hackers and me. Thus any change now might break PGP - which is still widely used (although mostly for encryption).

We suggest the use of the keyboxd for a reason. The use of multiple keyrings has always been a problem and has been kept on demand from a couple of people. Eventually things change and for a new installation the use of the keyboxd is the suggested way to run GnuPG. Support for pubring.gpg and even pubring.kbx may eventually be removed - not now or in the next year but it may happen. You have been warned ;-)