@michaellarabel A transactional workload like pgbench's tpc-b like workload, be it readonly or read-write, will never use JIT. You're just seeing run-to-run variation.
https://www.phoronix.com/scan.php?page=news_item&px=PostgreSQL-11-JIT-Benchmarks …
The overhead of doing JIT for queries as short as those used in that benchmark is never worth it. Even leaving the overhead of doing JIT aside, the bottlenecks are entirely elsewhere. The planner realizes that and avoids the use of JIT.