Describe the bug

The TechEmpower benchmark runs for rack-puma-mri-sequel-raw and rack-puma-jruby-sequel-raw show a significant performance regression after upgrading to Puma 7.0.0.pre1.

With Puma 6.6.0:

https://www.techempower.com/benchmarks/#section=test&runid=9d2e8e26-cb0c-48ff-8b97-dc2081b2db61&w=zik0zj-zik0zj-zik0xr-8vn08v&f=zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zf-zik0zj-zik0zj-zik0zj-zik0zj-18y67

After upgrading to 7.0.0.pre1

Note that rack-sequel-postgres-puma-mri and rack-sequel-puma-mri still run on Puma 6.6.0.

https://www.techempower.com/benchmarks/#section=test&runid=5db88eea-ccd9-4d03-bcd0-e5c8f90b921d&w=zik0zj-zik0zj-zik0xr-zik0zj&f=zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zj-zik0zf-zik0zj-zik0zj-zik0zj-zik0zj-27wr27

Puma config:

The config can be found in the repo.
The following command is used for starting puma.

bundle exec puma -C config/puma.rb -b tcp://0.0.0.0:8080 -e production

To Reproduce

Checkout the repo at: https://github.com/TechEmpower/FrameworkBenchmarks

Run the rack test with:

./tfb --test rack

I'm seeing things like:

rack: [8] ! Terminating timed out worker (Worker 4 failed to check in within 60 seconds): 29
rack: [8] - Worker 4 (PID: 54670) booted in 0.01s, phase: 0
rack: [8] ! Terminating timed out worker (Worker 7 failed to check in within 60 seconds): 41
rack: [8] - Worker 7 (PID: 54724) booted in 0.01s, phase: 0

The results can be expected with jq:

jq '.rawData.fortune.rack[4].totalRequests' results/20250829094943/results.json

Expected behavior

7.0.0.pre1 should perform better/equal to 6.6.0.