I guess what I'm getting at -- and could have communicated better -- is that this approach of "buffer/queue/cache/defer everything because it's faster right now, never mind that we might have to stall and flush the pipeline all at once later" has worked its way up the stack, with the worst-case stall growing in magnitude as the path cascades through progressively more layers and bigger buffers (cf. bufferbloat in networks [1]). Among other issues, it fools people into thinking that expensive operations are cheap. I think that pattern is far more responsible than language-imposed overhead for the weird quasi-random delays that are seen in modern systems. You can certainly connect this to popular idioms and some standard library facilities in some higher-level languages, but I think that connection is largely a historical accident rather than having much to do with the languages themselves.

[1] https://en.wikipedia.org/wiki/Bufferbloat