Another way to skin the same cat: Safety is not a binary variable even though we have this mental model of safe/unsafe. In fact safety is continuous ranging from high risk (unsafe) to low risk (safe).
With more information (more trials/subjects) we get a better estimate (with less uncertainty) of the associated risk. The properties (safety/risk) of the vaccine don't change. What changes is the uncertainty around our estimations.
This is still a simplification (e.g. there could be unknown unkowns) but I find it a better heuristic than safe/unsafe.
I agree on the simplification. No such thing as "safe", only "safe enough" which assumes every one's scale of safety is the same. That seems presumptuous.
With more information (more trials/subjects) we get a better estimate (with less uncertainty) of the associated risk. The properties (safety/risk) of the vaccine don't change. What changes is the uncertainty around our estimations.
This is still a simplification (e.g. there could be unknown unkowns) but I find it a better heuristic than safe/unsafe.