I'm not sure where that's the case, but it's certainly not the case with cryptography, where "clunky" means "mired in design decisions made before the era of authenticated cryptography".
To me, "old and clunky" also tends to imply "memory unsafe", or "written in Perl so old that a pipe filter in the wrong place in the input coughs up a shell", or "better make sure nobody's name is O'Connor because that includes SQL metadata", or "lol remember temp file races". But I'm prepared to concede that there may be places where the old stuff is better than the new, and eager to hear examples.
My 10Mbps hub (yes, not a switch) has been with me for decades. I've owned lots of smarter (switched, managed) and faster (100Mbps and then 1Gbps) devices that used less power (turns out when 10Mbps was invented nobody cared about "power saving"...) but over the years lots of those have dropped dead whereas the 10Mbps hub still works, which is why I still own it even though it isn't currently wired up, I know that sooner or later a 1Gbps switch will die and I'll hook that 10Mbps hub in there until a new one arrives.
Now, is it "better"? Overall, no. Or else I wouldn't keep buying Gigabit switches. But is it more robust? Yes. And if that's what you care about...
Thanks very much for the additional depth on what “old and clunky” could mean in cryptography. I am no expert in this area, and I agree that all these nightmare scenarios you mention (and many more, I’m sure) are very concerning. I prefer to use old (tested and tried) Unix/BSD tools for interactive development rather than their latest rewrites or modern GUIs. In drug design, I’d rather have a clearly understood 30-year-old assay with known failure modes rather than the latest promising experiment that may still have unknown failure modes. But generally speaking, I am a sucker for new tech and I seem to never learn to stop playing with new shiny toys. In a field with explosive growth field, like machine learning, at least I can use the benchmarks against the state of the art to help me decide how or when to advance technologies (every month or so, it seems). So.. is it worthwhile to learn to use age?
cperciva's spiped is, at least, deliberately old-fashioned; and I'd be inclined to trust it over most newer designs. (Wireguard does appear to be pretty awesome.)
I do agree that lots of legacy cryptosystems just aren't very good.
