parallel_tuple_cost
No restart — settable per sessioncost factor (per row shipped worker → leader)
The planner's per-row cost of moving a row from a worker to the leader.
What it does
After a parallel subtree runs, each row produced has to be shipped through the Gather to the leader. parallel_tuple_cost (default 0.1) is how much the planner charges per row for that transfer — which is why parallelism is more attractive for queries that aggregate or filter than for ones that ship all rows.
How to tune it
Rarely changed. Lowering it (with parallel_setup_cost) nudges the planner toward parallelism on row-shipping-heavy queries; raising it discourages it.