perf: add missing indexes on member_account_id and velocity account_id

[sandipndev]

Summary

Add two missing database indexes that are hit on every post_transaction_in_op call:

• idx_cala_account_set_member_accounts_member_id on cala_account_set_member_accounts(member_account_id)
• idx_cala_velocity_account_controls_account_id on cala_velocity_account_controls(account_id)

Why these matter

Two queries run on every single cala transaction post:

1. fetch_mappings_in_op — finds which account sets each account belongs to:

SELECT m.account_set_id, m.member_account_id
FROM cala_account_set_member_accounts m
JOIN cala_account_sets s ON m.account_set_id = s.id
WHERE m.member_account_id = ANY($2)

2. find_for_enforcement — finds velocity controls for accounts:

SELECT values, velocity_context_values
FROM cala_velocity_account_controls v
JOIN cala_accounts a ON v.account_id = a.id
WHERE account_id = ANY($1)

Without indexes, PostgreSQL does a sequential scan of the entire table for each query. With 1000 loans in lana-bank, cala_account_set_member_accounts has ~14,000 rows (14 accounts per loan). Every transaction post — including every interest accrual — scans all 14,000 rows to find the ~2-3 matching ones.

With the index, it's an index lookup — O(log n) instead of O(n). The benefit grows with data volume. At 10,000 loans with daily accruals, you'd have 140,000+ rows being scanned on every single transaction post.

Benchmark (lana-bank, 100 loans parallel=10)

Metric	Before	After	Change
100 loan creation	39.4s	35.3s	-10.4%
10 days interest accrual	28.9s	26.6s	-8.0%

Test plan

• Existing cala tests pass
• lana-bank e2e tests pass with local cala pointed at this branch

---

Found via autonomous performance investigation using autoresearch on the lana-bank loan workflow

🤖 Generated with Claude Code

[github-actions]

📊 Performance Report

Commit: 7add0e8
Updated: 2026-04-02 15:15:56 UTC

Cala Performance Benchmark Results (non-representative)

Criterion Benchmark Results (single-threaded)

Benchmark	Time per Run	Throughput	% vs Baseline
post_simple_transaction	5.781ms	172 tx/s	0 (baseline)
post_multi_layer_transaction	6.584ms	151 tx/s	-13.0%
post_simple_transaction_with_effective_balances	8.220ms	121 tx/s	-42.0%
post_simple_transaction_with_skipped_velocity	6.849ms	145 tx/s	-18.0%
post_simple_transaction_with_velocity	7.932ms	126 tx/s	-37.0%
post_simple_transaction_with_hit_velocity	3.701ms	270 tx/s	+35.0%
post_simple_transaction_with_one_account_set	7.015ms	142 tx/s	-21.0%
post_simple_transaction_with_five_account_sets	7.072ms	141 tx/s	-22.0%

Load Testing Results (parallel-execution)

Scenario	tx/s
1 parallel	103.89
2 parallel	151.89
5 parallel	181.56
10 parallel	182.56
20 parallel	179.55
2 contention	144.21
5 contention	153.40
2 acct_sets	126.50
5 acct_sets	139.58

---

Note: Performance results may vary based on system resources and database state.

Last updated by commit 7add0e8