I don't see a reason for always executing the polyfill, even for browsers that support modulepreload. Looks like in the polyfill this is what is being done https://github.com/guybedford/es-module-shims/blob/main/src/es-module-shims.js#L318

But we already have support detection for modulepreload in importAnalysisBuild

const relList = document.createElement('link').relList
const supportsModulePreload = relList && relList.supports && relList.supports('modulepreload')

I added a guard to the polyfill so we don't change things for chromium where this is already supported.

We discussed with the team that this feature is going to be merged before the beta for the next minor (2.5)

@patak-js ist it possible to also to check polyfillModulePreload where skip const is defined e.g.
const skip = config.build.ssr || !config.build.polyfillModulePreload
otherwise it's looks like polyfill can't be removed?

@croban could you raise a bug report with a reproduction if you found a problem? AFAICS the polyfill plugin is not included in config.build.polyfillModulePreload is false here

Did this never make it into 2.4.0 or later? I am currently running vite 2.6.7 and I see the modulepreload-polyfill script getting loaded in firefox, but I am also seeing some files downloaded twice in Firefox. Interestingly, the modulepreload-polyfill is being loaded using rel=modulepreload and is one of the files being downloaded twice.

Would you open an issue with a reproduction so we can better track this?