CDC resolved ts and checkpoint lag unstable and up to 2min

[fubinzh]

What did you do?

1. Deploy TiDB cluster with 2 CDC nodes (16c32g)
2. Create kafka simple protocol changefeed
3. Run workload for 4k tables (row width: 1.8k, cdc throughtput ~12k

What did you expect to see?

CDC lag should be stable

What did you see instead?

CDC resolve ts and checkpoint lag are very unstable, up to 2min sometimes.

Versions of the cluster

Release Version: v7.5.2-pr11031 Git Commit Hash: dd48f9e5bbd997a3885638315d5068bb194e2c80 Git Branch: heads/refs/tags/v7.5.2-pr11031 UTC Build Time: 2024-05-07 04:49:30 Go Version: go version go1.21.6 linux/amd64 Failpoint Build: false

[fubinzh]

/severity major

[fubinzh]

/label affects-7.5

[fubinzh]

/assign @hicqu

[asddongmen]

Duplicated with #11078