"who can forget the very first Ariane 5 launch..."
Indeed. A masterclass in fu**ked software engineering. But note also it was a case of "If it's not broke, don't fix it." In fact the SW should have been stripped out.
However 74 launches later they have not repeated it.
"SpaceX have done a good job so far, let down in one accident by a faulty component, and by pushing the envelope on rocket fueling in the other one."
Actually it's starting to look like they have had a recurring problem with COPV's (which BTW have historically had a pretty good safety record on all other LV's)
"All done on a budget a lot lot lot lot smaller than Ariane or Atlas had."
True, although it had long been suspected that industry cost models (which institutionalise mediocrity), government cost plus contracting and purchasing regulations (also "Just return" for European projects and the US practice of a contractor in every Congressional district), not to mention the "sub contract everything" meme multiplied the cost of such a project.
Surprise, surprise. The answer turned out to be yes.
The problem is the Aerospace Corporation have a thing called the "5/8" rule. A failure with in 5 launches of initial launch is probably a mfg flaw. At 8 or above it's likely a design flaw. This is the 2nd time a COPV issue has destroyed a payload (and it looks like the third serious incident they've had with COPV's).
That's looking to be a bit of a pattern.