You can adjust this, but the primary calculation is a practical limit on performance. This is controlled in limits.conf.
The maximum number of concurrent searches is calculated based on max_searches_per_cpu times the number of CPU cores in the system (as reported by the OS, which means VMs often lie when given multiple vCPUs but are running on a small number of hardware cores thanks to threading and shady hypervisor oversubscription models) + base_max_searches.
max_searches_per_cpu defaults to 1
base_max_searches defaults to 6
Therefore, in a reference system with 12 cores, you have 1 x 12 + 6 = 18.
If you need to crank this up, you can linearly scale this by core count using max_searches_per_cpu by setting it to 2 or more. This changes the math to be 2 x 12 + 6 = 30 or 3 x 12 + 6 = 42. If you just want to tweak it up by a fixed amount, adjust base_max_searches to a higher value.
All of the above comes with the caveat that generally these are bad ideas to implement in a production setting unless you have a highly underutilized system performing large numbers of extremely low memory and low CPU usage searches.
