`TestDeltaCorrectness/delta-006-007` much slower than the others

[lmbarros]

The test case delta-006-007 (part of the TestDeltaCorrectness integration test) runs much slower than its sibling test cases, which do the same thing with different data. I don't why, but I don't think the different input data alone could explain the difference. Maybe we are hitting some performance corner case there?

--- PASS: TestDeltaCorrectness (28.07s)
    --- PASS: TestDeltaCorrectness/delta-000-001 (0.38s)
    --- PASS: TestDeltaCorrectness/delta-000-002 (0.50s)
    --- PASS: TestDeltaCorrectness/delta-001-003 (1.20s)
    --- PASS: TestDeltaCorrectness/delta-000-003 (0.99s)
    --- PASS: TestDeltaCorrectness/delta-004-005 (1.13s)
    --- PASS: TestDeltaCorrectness/delta-006-007 (15.43s)
    --- PASS: TestDeltaCorrectness/delta-006-008 (2.06s)
    --- PASS: TestDeltaCorrectness/delta-008-006 (1.89s)
    --- PASS: TestDeltaCorrectness/delta-010-011 (2.45s)
    --- PASS: TestDeltaCorrectness/delta-010-012 (1.93s)