[PITR] Multiple PITR restores on sequences in conjunction to usage of sequences in a non_PITR database during restores generates core files in the cluster, brings master down and YSQL shell becomes unresponsive

[kripasreenivasan]

Jira Link: DB-593

Description

Steps to reproduce this issue:

1. Create a database db_1
2. Setup PITR snapshot schedule on db_1
3. Create table1, table2 with serial and bigserial sequence columns
4. Note down time
5. Alter table 1 to add a Foreign key
6. Restore to time in 4
7. Alter table 1 to add primary key
8. Restore to time in 4
9. Keep executing DML and DDL commands on the tables and restore. I executed restore about 18 times.
10. Create a new database db_2 and switch console to use the new database. Do not setup PITR on db_2
11. Note down time
12. Create a sequence in db_2, create a table using the sequence, populate data. Simultaneously, create a new sequence in db_1 and create a table to use the sequence in db1, populate data
13. Restore to time in 11
14. Perform few more DMLs in the table in db_2
    Issue:

• There are a lot of core files generated on the cluster.
• YSQL shell becomes unresponsive.
• Master goes down

Attaching the terminal dumps for exact commands executed during table creation
non_PITR_db_shell_01.txt

PITR_db_002.txt
yb-admin.txt

CC: @lingamsandeep , @bmatican