Critical appraisal: Failure to reject null hypothesis for primary endpoint

Superiority trial:
Why is it inappropriate to interpret statistically significant secondary endpoints when the primary endpoint failed to attain statistical significance? I understand the value that directional consistency between primary and secondary endpoints can have in adding credibility to the secondary endpoints (e.g. making it less likely that the observation is due to chance alone). But why can't multiplicity adjustments save the day? I know they cant, but I need the "why".

Non-inferiority trial:
I find it difficult to wrap my head around the fact that statistically significant secondary endpoints are more credible if the primary endpoint attained non-inferiority (i.e. rejected the null-hypothesis of inferiority). At the end of the day, the primary & secondary outcomes are still not directionally consistent, so why is the credibility of the secondary outcome better in this case? Is it simply a matter of "because the null was rejected for both in this non-inferiority example???

I really appreciate any help thinking this through. Also let me know if you think I've made some dangerous assumptions.