Clvmd has been consuming a lot of memory resource to process every single request. Each request in current architecture invokes a thread, and by default such thread is allocating ~9MB of RAM for stack. Since clmvd uses mlockall to keep all pages in memory, 200 parallel request easily grown running binary size to 2GB leading either to OOM killer or very slow system.
Fix has reduced the default thread's stack size to 128KB which is enough for current version of lvm to handle all tasks. This leads to massive reduction of memory used runtime by clvmd.