Multicollinearity - Variance Inflation Factors

#1
Hi I am doing a linear regression coursework project and am having some difficulties I hope someone can help me with.

I am analysing the VIFs for the full regression model with all the explanatory variables and have found 2 VIFs, those for the variables log(HCPot) and log(NOxPot), that are above 10 hence I have concluded there is some correlation between these variables, producing a scatterplot of log(HCPot) vs. log (NOxPot) confirms this.

However what I cannot understand is despite this obvious multicollinearity the P-value for log(NOxPot) is still 0.044 suggesting it is statistically significant. In fact when I remove log(HCPot) from the model the p-value for log(NOxPot) becomes 0.116 hence suggesting it is no longer important. Can anyone help me interpret what this means?

What is more baffling is there is another variable log(SO2Pot) which has a VIF less than 5. When I do a full regression but remove log(HCPot) AND log(SO2Pot) the p-value for log(NOxPot) becomes 0.035 suggesting it is now significant?

Can anyone help me interpret these results?

Thanks.