Bug: Island-based evolution not distributing programs across islands

[eilamt]

Description

When using island-based evolution with num_islands > 1, all evolved programs are placed in Island 0 instead of being distributed across the configured islands. This effectively disables the island-based evolution feature.

Steps to Reproduce

1. Run the function_minimization example exactly as documented:

cd examples/function_minimization
python ../../openevolve-run.py initial_program.py evaluator.py --config config.yaml

After completion, check the checkpoint metadata:

cat openevolve_output/checkpoints/checkpoint_10/metadata.json | python -c "import json,sys; d=json.load(sys.stdin); print('Island generations:', d['island_generations']); print('Island sizes:', [len(i) for i in d['islands']])"

Expected Behavior
With num_islands: 3 configured, programs should be distributed across all 3 islands with roughly equal generations per island.

Actual Behavior
All programs go to Island 0. Islands 1 and 2 remain empty:

"island_generations": [10, 0, 0]
"islands": [[...11 programs...], [], []]
"island_best_programs": ["...", null, null]

Root Cause Analysis
In openevolve/process_parallel.py, the iteration worker determines a target island using round-robin (next_island()), but when the child program is added to the database via database.add(), no target_island parameter is passed. The database.add() method then defaults to using the parent's island, which is always Island 0 since:

The initial program is seeded to Island 0
When sampling from an empty island, sample_from_island() falls back to sampling from any available program (which is in Island 0)
Children inherit their parent's island
This creates a feedback loop where all programs end up in Island 0.

Environment
OpenEvolve version: latest main branch (commit ad9c9c1)
Python: 3.11
OS: macOS
Additional Context
This bug affects all examples that use num_islands > 1. Tested with both function_minimization (3 islands) and a custom graph coloring example (5 islands) - same behavior observed.

[codelion]

I will take a look at this, thanks for reporting.

[eilamt]

awesome!