This question comes from my previous question asked here
$$\vec v_2 = \vec v_1 + \vec v_\perp.$$
I think the logic I'm making a mistake is I presume that $\vec v_\perp = adt$ in x-direction. This presumes that during $dt$ time interval, $a$ is constant, hence this gives us > 0 value which means $\vec v_2 > \vec v_1$.
After seeing another question, I realize $a$ is not constant during $dt$ but even if it's not and changing, then we could split $dt$ again in even smaller intervals($dt_1, dt_2, dt_3,...$) and say that $v_\perp = a_1dt_1 + a_2dt_2 + a_3dt_3+...$. This way, $a$ is not constant over $dt$. but now this should give us 0 and the only way this is possible is if over small $dt$ interval, $a$ increases, then decreases and overally results in 0. What's the non-intuitive, solid proof that this is what happens?
Another way proving this is $$ v_2 = v_1\sqrt{1 + (\frac{adt}{v_1})^2} = v_1(1+\frac{1}{2}(\frac{adt}{v_1})^2 - \frac{1}{8}(\frac{adt}{v_1})^4 + \frac{1}{16}(\frac{adt}{v_1})^6)$$ and if $dt -> 0$, this becomes $v_2 = v_1$, but I don't understand this - what we gain with $dt -> 0$ intuitively? even if we take the interval that approaches zero, it doesn't say anything about acceleration and making the interval 0 doesn't seem proof to me.