Harm ratio = harm caused by vaccine / harm caused by virus
Exposure ratio = projected people who would be infected by the virus / projected people who would take the vaccine
In an airborne, high-R_0 pandemic situation, the exposure ratio is high enough that only one of two cases apply:
1) The potential health problems caused by small-scale exposure to the spike protein alone is relatively small compared to problems caused by infection, in which case vaccinate everyone. This is the current situation.
2) The potential health problems caused by small-scale exposure to the spike protein are significant (for example, hypothetical antibody cross-reactivity with a normal human protein, as mentioned in a parallel comment), in which case obviously don't vaccinate everyone, but lock down the world for a few weeks since having much of the world suffer autoimmune disease is untenable.
The potential health problems are unknown. For example, we are assuming that Myocarditis/Pericarditis is extremely rare, or at least more rare than in a viral infection. Do we really know that, for all age groups and all vaccines? I have doubts, we are still in the discovery phase:
The spike protein is not necessarily a problem, but both NLP and viral vectors have never been rolled at the current scale. TTS seems to be particular to viral vectors, Myocarditis/Pericarditis more so to NLP.
Pushing out a vaccine that causes harm, even if it's substantially less than the virus it protects against, is asking for a lot of trouble.