cmp-buffer

Indexing large files can consume tens of megabytes of RAM, as acknowledged in the README, requiring manual configuration to avoid performance hits.

The documentation warns about incorrectly setting options like keyword_pattern outside the options table, which can lead to setup errors and confusion.

Only provides word-based completion from buffers, lacking code context understanding compared to LSP sources, limiting usefulness for complex coding tasks.

While indexing is async, watching for changes is synchronous, which might cause lag during large edits like pasting multi-line snippets, as noted in the indexing section.