I don't think the deprecation was really intended for node core as much as it was for userland. See the original discussion here. One of the reasons is that Object.assign() is/was slower. The changes would need to be benchmarked first before this PR would be landed, to see if it is still slower and by how much.

quick & dirty benchmarking shows that Object.assign({}, process.env) is still about 50 % slower for me than util._extend({}, process.env), so yes, this PR would probably not improve things performance-wise.

I'm going to close this for now. We can revisit when performance of Object.assign() improves, or if it is used in a way that makes it a better option than util._extend().