Since we figured out that is not straightforward how could we write a regression test for WT-15158 it was decided to merge a quick fix first separately. That ticket scope is to create a regression test that would reproduce the same problem if we rollback the fix. 

Currently the preliminary idea of test is:

1. Create two tables: A and B
2. Create a follower
3. Put some data on the leader to A
4. Checkpoint #1
5. Pickup the checkpoint #1 on the follower
6. Open a cursor on the follower for A
   1. The idea is that it should make the checkpoint #1 being marked as "inuse" for follower and avoid deleting it from ckpt_track structure
7. Put more data on the leader to A
8. Checkpoint #2
9. Pickup the checkpoint #2 on the follower
10. Open a cursor on the follower for B
    1. The idea is that before the fix it should assign `prune_timestamp` to Checkpoint #2 while the Checkpoint #1 is still in use.
11. Put more data on the leader to B
12. Checkpoint #3
13. Pickup the checkpoint #3 on the follower
    1. The idea is that according to the current logic it should try to assign timestamp from the checkpoint #1 since it's still in use on the follower to the prune_timestamp for B which is set to the timestamp for the checkpoint #2 already.
    2.  

The problem is that currently the 6th step doesn't cause the table to be marked as "inuse" or even "found" for later pickups.