Indexing memory limit flags ignored in v1.20

In Meilisearch v1.20, the --max-indexing-memory CLI flag and MEILI_MAX_INDEXING_MEMORY environment variable are not properly enforced. Users report no observable difference between 1MiB, 1GiB, and 4GiB limits during indexing of large datasets (2 million documents). Memory consumption remains unconstrained regardless of configuration.

Upgrade to v1.21+ where the issue appears addressed. For v1.20, combine memory limit settings with --experimental-reduce-indexing-memory-usage flag and reduced thread count. Use Docker memory constraints as a hard limit to prevent runaway consumption. Verify settings are applied by checking actual memory usage during indexing.