Because running 100s of JVM's, ie one per shard and multiple shards per node, is far more of a waste compared to the ability to tune one single shard's JVM.
If you are having problems, then perhaps you can share details on that aspect.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.