Phase one is about dosage at least as much as safety. The most likely reason to fail is probably that it doesn't produce an immune response at any reasonable dosage.

Is there any risk of a "doesn't produce immune response" failure that would make it unsafe to start giving to healthcare workers immediately? Could it make later vaccines ineffective?

I'm sure that it would ruin the data for later vaccines, but if high-risk people just take each new vaccine as it starts trials, their data doesn't have to count towards the actual trial.

Don't downplay the safety. Phase 1 can fail because of lack of immune response OR safety concerns. Phase 2 and 3 trials are also about safety, and both of them somewhat often fail because of safety issues.

A vaccine can make a disease worse by priming the immune system to overreact and attack the entire body when a single cell is infected.

The more we know about a vaccine the more confident we are. The flu vaccine is changed every year without long safety studies: we have learned enough about how people response to the flu vaccines in use to not need them anymore. When the same process is used to make a non-flu vaccine though we still have to start over because there isn't any way to be sure.

You are quickly getting beyond my area of knowledge though. I didn't answer all of your questions because I don't know.

A complicated example of your point about "priming the immune system to overreact" is

https://en.wikipedia.org/wiki/Dengvaxia_controversy

Dengue fever is a strange and terrible disease where -- unlike most infectious diseases! -- contracting it once can make future infections worse through autoimmune mechanisms (if I remember correctly, you typically become immune to the particular strain that caused the first infection, but may become more susceptible to severe reactions from a later infection involving a different strain). Apparently the Dengvaxia vaccine, while effective at reducing the chance of dengue infections, also has an effect similar to surviving an infection: it increases the chance that an infection will result in severe disease. That makes it a more complicated question whether it's a good health intervention in particular circumstances, and in the Philippines there was a big scandal where a childhood immunization program using it may have reduced the total number of people who contracted dengue, but also increased the severity of the worst cases.

To my knowledge the pathogens that cause COVID-19 and dengue fever are extremely different, but it seems logically possible that there could be a COVID vaccine that greatly reduced people's chance of getting sick at all, while also increasing the chances of severe symptoms for those people who did get sick.

In therory, the mRNA vaccines should be extremely safe, as they are so specific and only require a carrier but no other substances in the vaccine. In practise, you still need to test. There is always a small risk of something unexpected happening. Like the vaccine from Australia which has been completely abandoned - not sure whether it also was mRNA or some other kind of vaccine. It didn't cause any health problem, but the vaccinated people were tested as HIV-positive. Of course they were not, but the vaccination caused those false positive. So while no direct heahlt impact, it is quite a disaster if one couldn't test people for HIV any more.

> So while no direct heahlt impact, it is quite a disaster if one couldn't test people for HIV any more.

It only invalidated certain types of HIV tests. The impact would be that some subset of existing HIV tests would have to be abandoned. That would impose a cost on the health system.

The real issue is that there were other vaccines being developed that were equally effective but which did not have this drawback. If this was the only vaccine available, the health system would have worn the burden of changing HIV test regimes; given several other effective vaccines are becoming available, it wasn't clear that the cost of imposing massive changes on HIV testing regimes was justified.

There are three options, the mRNA "carrier", such as the lipid particle, the mRNA itself, and of course the protein product.

The first part is hard to guess without knowing what's in those delivery methods.

mRNA, or RNA in general, can form secondary structures which in theory could affect the the cell machinery. Depending on the details this could produce side effects to the cells that take up the mRNA. This would probably be localized and a strong secondary structure may inhibit translation also, rendering the vaccine ineffective. This can be checked in silico and in vitro during design to some certainty.

The protein product I would guess is the most likely to cause problems. Essentially you can encode any protein in mRNA, so you could make the cells that take it in produce snake poison, for example, if you wanted to. Depending on how the virus protein is modified, and how it reacts with the cells/body when present apart from the rest of the virus it assembles with could bring all kinds of side effects. The severity probably depends on what exactly it interacts with, and if it's only local or it gets into the bloodstream etc.

Much of the danger of vaccine development is the possibility of activating antibody response to an off-target cell that expresses the same feature the vaccine is expressing. You can't really find those reliably without human trials.