CDC cloud: CDC OOM when upstream has historical data and has 60K tables

[Tammyxia]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. What did you do? If possible, provide a recipe for reproducing the error.

• There are 8 CDC servers, each has vcpu36c mem72g , 1 changefeed, 6w tables.
• Stop changefeed, add workload in upstream tidb, data size less then 40GB.
• Resume changefeed. Meanwhile, workload IO in upstream is still running...

2. What did you expect to see?

• No any error.

3. What did you see instead?

• CDC OOM persistently...

4. Versions of the cluster

   • Upstream TiDB cluster version (execute SELECT tidb_version(); in a MySQL client):

     4.0.14
     

   • TiCDC version (execute cdc version):

Release Version: v4.0.14-42-gc24e30f
Git Commit Hash: c24e30f
Git Branch: HEAD
UTC Build Time: 2021-08-11 06:03:32
Go Version: go version go1.16.4 linux/amd64
Failpoint Build: false
```