I have a saying that isn't perfectly true but often will apply to "fixes" like this -
If you don't know why the fix worked, you may not have actually fixed it.
If you don't know why the fix worked, you may not have actually fixed it.
The best they could do was work around it.
Sometimes workarounds are the best you can do until your vendor provides a real fix.
If your system doesn't work, and you just plonk around at values, until, very surprisingly, the system starts behaving well and you the call it working... well it might be working now. But it's just accidental correctness. As soon as something causes the system to bank left, something's gonna break and no one knows how to fix it - and you're back to square one.
On the other hand, as hard as it is, if you can clearly tell why your fix will restore function to the system without even applying it, you have deliberate correctness and function. If done right, it is very boring, because exactly and only the expected thing will happen. You should know about the unknowns and plan around those as well, so even if an unknown bites you, it's a known and handled unknown. This can be exhausting to make happen, because it is much harder, but those systems will just work.
But this is a fight I have with some development teams probably forever. "But we poked at the values, and that stopped the flames. It is fixed!" "but why?" "Dunny. But no fire anymore. All good." And then 2 weeks pass, and there is more fire and everyone is like "Oh but why would this happen? How should we have known for this to happen again"
For example, the article doesn't get to a root cause in an absolute way. There is no absolute SEGFAULT of the OS causing the misbehavior. However, they nail down the crash to a gif, and if the gif is in, it crashes, and if the gif is out it doesn't. If the gif is loaded otherwise it crashes, too. At that level, to me, that would be enough, because we're users of the browser's rendering there.
Finding a solid cause that can demonstrate and reproduce a problem, and basing a workaround around that at a boundary you're unwilling to cross can be fine. If it's within the company, it absolutely is fine as long as you escalate beyond that boundary.
However, I have enough teams who are like "Oh, we set all values to 25 one by one and when we arrived at flum-value at 25 it stopped crashing. Fixed." Why 25? Who knows. Why flum? Who knows. Maybe the other value changed at the same time fixed it? Who knows. Do we use 26 once it starts crashing again? Fuck knows. Maybe 24 is better?
We have no explanation for 25, so why would 25 be a good fix?
I hypothesize that Chrome simply has a global (i.e. cross-tab) per-toplevel-origin limit to the number of allocated accelerated drawing canvases it's willing to allow; and that when you go over it, Chrome forcibly de-allocates all the existing drawing canvases used by other tabs that have that toplevel origin loaded, thereby causing them to crash. It's probably a measure designed to prevent a site from from DoSing your computer by just allocating an infinite number of canvases.
I can also extend my speculation about all the times I've personally seen "Error 5", although this is probably dipping into superstition territory: I think the per-toplevel-origin draw-context limit isn't static, but is influenced by the amount of VRAM available on the computer as a whole.
So on an extremely VRAM-constrained system like a Raspberry Pi (or on a system that's running Cyberpunk 2077 on one display and Chrome on another, sharing a GPU) the per-toplevel-origin limit on accel draw contexts might be dropped as low as, say, 5: an amount low enough that just having one extra .gif on the single tab of the site you have open might be enough to make the tab fall over.
I suppose since they dont know the root cause it’s impossible to say. But I think the saying would fit better if they kept the gif but made some change that seemed to fix it without knowing why.