Thank you for quick response)

I couldn't build and test 2.1.9 right now, but bug is still here in 1.4.16 and
2.0.22.

I've created test script for this case:
https://gist.github.com/anton-ryzhov/a0dcfcaabe18fc6ad35e

Run ./gen.sh in some working folder and then try ./runtest.sh several times,
expire different subkeys, compare the result.

Do you need more information, or you can confirm and reproduce bug with given
description?

I'm using Linux Mint 12 Lisa, and I've tested on built-in 1.4.11 and on custom
built latest revision in repository - d64aa7.

1. I've created key with Primary key (P0), and 3 subkeys (S1, S2, S3). Export

this key for further tests.

2. Change expiration date of first subkey (S1). Everything seems OK.
3. Export whole key, remove it from gpg, import again - Everything is OK.
4. Back to step 2 - remove key, import original one.
5. Change expiration date of second or third subkey (S2, S3). Everything seems

OK again.

7. Export whole key, remove it from gpg, import again - we've missed S1 subkey,

and expiration date of changed subkey left as in step 1.

I've analyzed changes on each step via gpgsplit. My conclusion: GPG always edit
S1 subkey signature. Editing non-first subkey (S2, S3, S4…) edits (breaks) S1
signature. S2¸ S3… signatures leaved unchanged. GPG checks subkey signature
only at import. User doesn't know about subkeys breakage until he reexport it.