Slow Query Diagnosis from EXPLAIN

You diagnose slow SQL from an execution plan. Do not guess from the query text alone.

QUERY:
{paste}

EXPLAIN (ANALYZE, BUFFERS) OUTPUT:
{paste}

TABLE SIZES / EXISTING INDEXES:
{paste, or say unknown}

Return:

1. WHERE THE TIME GOES — the two or three nodes that dominate actual time. Quote the node lines. Ignore nodes that look scary but cost nothing.

2. ROW ESTIMATE ERRORS — every node where estimated and actual rows differ by more than 10x. This is usually the real cause; a wrong estimate makes the planner choose a wrong join.

3. ROOT CAUSE — one sentence. Not a list.

4. FIXES, in order of what to try first:
   - what to change
   - why it addresses the root cause above
   - what it costs (index size, write slowdown, maintenance)
   - how to confirm it worked, as a measurement not a feeling

5. WHAT NOT TO DO — the tempting fixes that will not help here, and why. Adding an index to a column that is already index-scanned is the most common one.

If the plan does not show enough to be sure, say what to re-run with and stop. A confident wrong diagnosis costs more than an honest gap.