Making commercial vendors who rely on open source software liable for bugs is fantastic news, that's how it always should have been. You can't have a commercial company throw their hands up and say "well github.com/cutefuzzypuppy is at fault for writing an open-source npm package we used so harm to our customers is not our fault!"
The biggest issue I see with this law is around liability for open source projects that people are using directly. It'll be disastrous if all open source software ceases to exist or be available in Europe because volunteers face legal liability if their code has a bug. In theory this could even impact people outside of Europe if they don't prohibit access to their code by EU citizens.
I release a lot of code on github. Most of it is just random crap that I wrote to solve a specific need or to explore an idea, and I put it up under an open source license because why not? If it helps someone, that's great. Now I need to be concerned that the random "example-service" project I wrote in C and published a decade ago to go with a blog post I wrote will end up costing me all the money I have ever or will ever earn in my career.
As far as
> when faced with a choice between being liable for their own code or being liable for open source code, most companies will choose to write their own code. If so, that would be a net harm to open source and user freedom
goes, even if that is true (I'm not really convinced) it doesn't really matter. What matters is finding the correct answer to "who is responsible" to which the answer can't be "nobody". And if it can't be nobody, then it must be somebody. And if it must be somebody, it absolutely shouldn't be some random guy who never specifically signed off on your usage of their open source code.
I think that's more-or-less fine. There's a concern that companies don't want to be responsible for open source code, and will write everything in-house instead. I wouldn't be surprised if some companies do that, even if it's a bad idea. I don't know how common it'll be, but the worst case scenario is that it turns out to be bad for developers and for free software.
The second, murkier issue, is what happens when there is no selling involved at all. If I download a debian iso, or clone some random repository on github, then there has been concern that the author of that code will be financially liable for any errors in the software. That would be very, very bad. Early versions of the law seem to explicitly say that it would be the case. More recent versions seem like they might have an exception so long as there is absolutely no money changing hands. It's unclear what would happen in cases where open source software accepts donations. It could still end up being harmful to individual developers and to open source software in general. It's hard to say.
Which would in turn be very bad for society.