Workflow [PR], commit [4f7f092]

Summary: ✅

Maybe just removing them from buzz tests is enough?

It's not the greatest feature for sure, but some people use it.

I understand that looking / fixing new buzztests report about that feature is not the greatest time investment, but having it uncovered by tests at all - does not sound nice. So maybe we can keep those happy path tests exists, to not brake at least those which exist.

Many of these tests are problematic. For example, assertion happens inside ZeroCopyLock.
But read this:

/** Caveats: usage of locks in ZooKeeper is incorrect in 99% of cases,
  *  and highlights your poor understanding of distributed systems.
  *
  * It's only correct if all the operations that are performed under lock
  *  are atomically checking that the lock still holds
  *  or if we ensure that these operations will be undone if lock is lost
  *  (due to ZooKeeper session loss) that's very difficult to achieve.
  *
  * It's Ok if every operation that we perform under lock is actually operation in ZooKeeper.
  *
  * In 1% of cases when you can correctly use Lock, the logic is complex enough, so you don't need this class.
  *
  * TLDR: Don't use this code if you are not sure. We only have a few cases of it's usage.
  */
class ZooKeeperLock

And here are all failures in master:

Have you left this on purpose?

• 02432_s3_parallel_parts_cleanup
• test_remote_blobs_naming/test_backward_compatibility.py
• test_encrypted_disk_replication/test.py

About 02432_s3_parallel_parts_cleanup - there were some complications, didn't have time to look.
Other tests remain not on purpose - please help remove zero-copy replication from them.

I've removed stateless tests here - #86006

And decided to keep integration tests, likely they are will not cause troubles anywya