seems to do is eventually force quit the tasks and not use the configured kill_signal. Will I need to drain the node and then individually loop through all allocations and send the proper signal with nomad alloc signal?
The node drain command is the recommended way to drain a node gracefully of all current work and mark it as ineligible for more work. The running allocations should be signalled to gracefully shutdown using the task kill_signal parameter. If they have not exited by the time the deadline is reached, then they will sent SIGKILL signal.
I do wonder if this PR and the linked issues describe the problem you are seeing? If so, this is something we are working on fixing.