The comments on the article also state: "With regards to time taken and peak working set, that was not our main focus. Our main focus was to reduce the number of allocations and reduce the time spent in GC. Unfortunately this data import process is not a standalone one time application. It is a part of a much larger application which is core to our entire platform."
It's possible that this has no meaningful impact to their application, but it's far more likely that the blog author just hasn't provided you all of the information you need to understand why it's important. I feel reading charitably is the best case here.
It's just that the headline sales pitch of the post is "from 7GB to 32KB". Unfortunately, that doesn't mean much and it's misleading. As is the title "Strings are evil". Those sentences combined with no obvious reason to ignore the primary optimization metrics and instead focus on something that doesn't affect either memory usage or run time very much, those things together make me feel like it's more likely the author either wanted a good blog post, or doesn't have a ton of experience with optimization. Those are both great, but that's no reason for me not to comment and try to add some clarity.
To me it seems a bit strange to say that time taken was not the main focus, and follow that with "reduce the time spent in GC." The singular primary reason to reduce allocations is to reduce run time of the program, there is no other reason here.
FWIW, it's highly likely that this program's run time could be further improved by more than 2x by allocating a single buffer the size of the file and doing a single fread() call, then processing the entire file in memory. That would increase the working set from 16MB to 300MB, or whatever the file size is, but would allow much faster processing than line buffered input.
I love Michael Abrash's introduction to the philosophy of optimization in the Black Book of Graphics:
"Before we can create high-performance code, we must understand what high performance is. The objective (not always attained) in creating high-performance software is to make the software able to carry out its appointed tasks so rapidly that it responds instantaneously, as far as the user is concerned. In other words, high-performance code should ideally run so fast that any further improvement in the code would be pointless.
Notice that the above definition most emphatically does not say anything about making the software as fast as possible. It also does not say anything about using assembly language, or an optimizing compiler, or, for that matter, a compiler at all. It also doesn’t say anything about how the code was designed and written. What it does say is that high-performance code shouldn’t get in the user’s way—and that’s all."
http://www.jagregory.com/abrash-black-book/#introduction