Closed syncmers have a window guarantee but less optimal spread. They seem to perform roughly equal (open syncmers slightly better) when benchmarked across some datasets. However, the benchmarks were done with parameters optimized for open syncmers. Also, the window guarantee helps ruling out 'seed deserts' as an issue for future development and improvement in various aspects.
It would be nice to have support for both open syncmers (current) and closed syncmers (implemented here with the same density: https://github.com/ksahlin/strobealign/commit/1aba24a5fd271ebd6b8d4db022f56a1ae9755c73).
Closed syncmers have a window guarantee but less optimal spread. They seem to perform roughly equal (open syncmers slightly better) when benchmarked across some datasets. However, the benchmarks were done with parameters optimized for open syncmers. Also, the window guarantee helps ruling out 'seed deserts' as an issue for future development and improvement in various aspects.