Kleopatra behaves as intended (by me). Only subkeys meeting the following conditions are extended together with the primary key:

• skip revoked subkeys which would anyway be ignored by gpg;
• also skip subkeys without explicit expiration because they inherit the primary key's expiration;
• include all subkeys that are not yet expired or that expired around the same time as the primary key

The reasoning for the third condition is that "long" expired (encryption) subkeys should not be revived if the primary key is extended. Currently, "expired around the same time as the primary key" means up to +/-10 seconds different expiration dates. The normal use case is primary key with not expiring subkey (which is or was the standard result of creating a key; this is covered by condition 2) or primary key and subkey with same expiration date (which also is or was the standard result of creating a key; this is covered by condition 3).

If somebody adds a few more subkeys (e.g. a separate authentication subkey) just after creating the primary key then they may expect that those subkeys are extended together with the primary key. This would be an argument for allowing a greater expiration date difference, e.g. +/-1 hour.

On the other hand, the only truly transparent solution is a checkable list of all subkeys, so that the users can explicitly tell Kleopatra what they want and Kleopatra doesn't need to guess.

As in my test case Kleopatra didn't check the box for "with subkeys" when the extension wouldn't work I propose as minimal solution:

• disable the checkbox if no key would be extended anyway + info why the box is not checkable (via tooltip)

In case some subkeys could be extended and some not the best solution is debatable...
Kleopatra could maybe extend the keys for which it is possible and show an info window listing the keys for which it wasn't done? Plus advice to go to the command line to extend them if that's wanted.

But like it is now one would only go into the subkey view after things start behaving unexpectedly. After all, "I told the program to extend the subkeys", why should I verify if it really did this?
If the box "extend validity for subkeys" could not be checked in case no keys will be extended, I would then go to the subkey details, if I wanted to extend them, too.

It seems I did not express myself clearly.
I would prefer it if the "with subkeys" checkbox could not be enabled in case it won't extend any subkeys anyway.
With a bonus tooltip.

Ah, sure that also makes total sense, I thought you wanted it disabled if it did not extend all subkeys.

The testcase now works, gpg4win-beta-50.