[BUG] Performance regression between Node 22.17.1 and 22.18.0?

Is there an existing issue for this?

• I have searched the existing issues

This issue exists in the latest npm version

• I am using the latest npm

Current Behavior

In a non-trivial project, running various npm scripts that I have see a significant slowdown when I switch my Node version from 22.17.1 to 22.18.0 (both using npm@11.5.2).

Initially thought that this might be a Node issue but what I believe that the regression surfaces when running npm run. For example, here's me running typescript checks using node --run using 22.17.1 and 22.18.0:

❯ hyperfine --warmup=3 "mise exec node@22.17.1 -- node --run types:renderer" "mise exec node@22.18.0 -- node --run types:renderer"

Benchmark 1: mise exec node@22.17.1 -- node --run types:renderer
  Time (mean ± σ):      7.198 s ±  0.191 s    [User: 13.740 s, System: 0.857 s]
  Range (min … max):    6.903 s …  7.543 s    10 runs
 
Benchmark 2: mise exec node@22.18.0 -- node --run types:renderer
  Time (mean ± σ):      7.324 s ±  0.215 s    [User: 13.979 s, System: 0.873 s]
  Range (min … max):    6.986 s …  7.706 s    10 runs
 
Summary
  mise exec node@22.17.1 -- node --run types:renderer ran
    1.02 ± 0.04 times faster than mise exec node@22.18.0 -- node --run types:renderer

Marginal difference between Node versions using node --run.

However, here's me doing the same thing but using npm run instead:

❯ hyperfine --warmup=3 "mise exec node@22.17.1 -- npm run types:renderer" "mise exec node@22.18.0 -- npm run types:renderer"

Benchmark 1: mise exec node@22.17.1 -- npm run types:renderer
  Time (mean ± σ):      7.461 s ±  0.112 s    [User: 14.177 s, System: 0.908 s]
  Range (min … max):    7.298 s …  7.629 s    10 runs
 
Benchmark 2: mise exec node@22.18.0 -- npm run types:renderer
  Time (mean ± σ):      8.862 s ±  0.352 s    [User: 14.150 s, System: 0.928 s]
  Range (min … max):    8.579 s …  9.750 s    10 runs
 
Summary
  mise exec node@22.17.1 -- npm run types:renderer ran
    1.19 ± 0.05 times faster than mise exec node@22.18.0 -- npm run types:renderer

I'm observing a 1.5 second average increase in execution time between 22.17.1 and 22.18.0 for a relatively straightforward script. The disparity becomes much more noticeable when running more complex npm scripts. For example, here's a script I have that runs some other scripts sequentially and in parallel using npm-run-all2:

❯ hyperfine --warmup=3 "mise exec node@22.17.1 -- npm run intl:translations" "mise exec node@22.18.0 -- npm run intl:translations"

Benchmark 1: mise exec node@22.17.1 -- npm run intl:translations
  Time (mean ± σ):      2.844 s ±  0.071 s    [User: 4.585 s, System: 0.768 s]
  Range (min … max):    2.767 s …  2.990 s    10 runs
 
Benchmark 2: mise exec node@22.18.0 -- npm run intl:translations
  Time (mean ± σ):     15.240 s ±  0.402 s    [User: 4.942 s, System: 0.862 s]
  Range (min … max):   14.692 s … 15.766 s    10 runs
 
Summary
  mise exec node@22.17.1 -- npm run intl:translations ran
    5.36 ± 0.19 times faster than mise exec node@22.18.0 -- npm run intl:translations

~3 seconds in Node 22.17.1 compared to ~15 seconds in Node 22.18.0!

Expected Behavior

Ideally there wouldn't be such a notable performance regression.

Steps To Reproduce

I don't have a set of concrete repro steps but here's the environment I'm working with:

• macOS 15.6.1 (Sequoia)
• npm 11.5.2
• Using Node 22.17.1 and 22.18.0

Environment

• npm: 11.5.2
• Node.js: 22.18.0
• OS Name: macOS
• System Model Name: MacBook Pro 2019
• npm config:

; "user" config from /Users/andrewchou/.npmrc

//registry.npmjs.org/:_authToken = (protected)

; node bin location = /Users/andrewchou/.local/share/mise/installs/node/22.18.0/bin/node
; node version = v22.18.0
; npm local prefix = /Users/andrewchou/GitHub/digidem/comapeo-desktop
; npm version = 11.5.2
; cwd = /Users/andrewchou/GitHub/digidem/comapeo-desktop
; HOME = /Users/andrewchou
; Run `npm config ls -l` to show all defaults.

Start by reproducing the reported hyperfine comparisons between Node 22.17.1 and 22.18.0 on macOS, using both node --run and npm run with the listed scripts. Compare the simpler types:renderer case with the intl:translations case and isolate the regression before changing code. Done means identifying a reproducible cause and preventing the performance loss without regressing the older version.