kvnemesis: closed timestamp regressions during test injected reproposals

[arulajmani]

Describe the problem

We've seen KVNemesis routinely fail due to closed timestamp regressions. For example: #118471 (comment)

We attempted to fix this in #121965. However, we still see this sometimes. @pav-kv has a theory about why: #118471 (comment).

Jira issue: CRDB-42546

Epic CRDB-37617

---

Note to L2 triagers: if you encounter an instance of this failure during triage, point to this issue and close the test flake with the X-duplicate label.

[blathers-crl]

Hi @arulajmani, please add branch-* labels to identify which branch(es) this C-bug affects.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}