01
How MVCC handles concurrent writes without locks
Covers tuple visibility, transaction snapshots, and the cost of dead tuple accumulation on write-heavy tables.
Internals
02
Index types and when the planner ignores them
B-tree, GIN, GiST, and BRIN compared — with real explain output showing when each is skipped despite existing.
Query planning
03
WAL and durability: what fsync actually guarantees
Traces a write from application layer through shared buffers, WAL, and checkpoint — with data safety implications at each step.
Advanced
04
Partition pruning and its limits under parameterised queries
Why partition elimination fails when conditions arrive at runtime — and how to structure schema to recover it.
Partitioning
05
pg_stat_statements as a diagnostic baseline
Reading call frequency, mean exec time, and shared block hits to isolate slow queries before running explain.
Monitoring
06
Logical replication: slot lag and subscriber delay
Diagnosing replication lag caused by long-running transactions on the publisher — and approaches that reduce it without schema changes.
Advanced
07
Autovacuum tuning for high-churn tables
Per-table storage parameters, cost delay settings, and identifying when autovacuum cannot keep pace with dead tuple growth.
Maintenance
08
Row-level security: policy evaluation order and performance
How RLS policies compose with existing indexes and where policy design pushes work into sequential scans unexpectedly.
Security
09
Connection pooling with PgBouncer: transaction vs session mode tradeoffs
Examines which PostgreSQL features break under transaction pooling and the operational differences at scale.
Infrastructure
10
CTE behaviour changes between PostgreSQL 11 and 12
The shift from CTE as optimisation fence to inlined subquery — with query plans showing the difference across both versions.
Query planning
