"dnf update" considered harmful

If those interruptions are just random then the functional tests might take a long time until they hit interesting cases. I think something smarter would be needed here. The same applies to out-of-memory/fd/etc conditions, for which dpkg is designed to be resistant against, and in practice it should currently be (barring bugs), in the same way it should be resistant against the above mentioned abrupt crashes.

I'd actually expect a stream of bug reports if any of the above would happen, as we've seen in the past, for both types of bugs (a recentish example of the latter could be the infamous ext4 zero-length problems).

In any case, those are indeed things that I'd like to (eventually) add as part of its test suite, but have not found the time yet. Take into account that when I joined the maintainer team at the time, around 10 years ago, dpkg didn't even have any kind of unit nor functional tests! Now it's a bit better, but it could certainly be improved:

In-project test suite: http://dpkg.alioth.debian.org/coverage/
External functional test: https://anonscm.debian.org/cgit/dpkg/dpkg-tests.git/ (need coverage reports for this one)

And if you are interested, of course patches more than welcome!