1- 50/50. At the point of inversion both probabilities are the same.
2-I think you are missing some key points here. You always need to include the energy structure. For example in a 4-level system: $E_0$ ground level, $E_1$ upper pump level, $E_2$ upper laser level, $E_3$ lower laser level. The inversion between ground state $E_0$ and upper-pump level $E_1$ is always close to 0. And the population of the lower laser level $E_3$ is always really low. This means that 1: you always have a high probability of exciting an electron from $E_0$ to $E_1$ because the occupancy of $E_1$ is low. And 2: that for virtually every electron decaying from $E_1$ to $E_2$ you achieve inversion. And after an electron falls back to $E_3$, it almost immediately goes down to $E_0$ again. This leaves the lower laser level practically free, and with lack of electrons to be excited back from $E_3$ to $E_2$.
What this entails is that you virtually achieve very high values of inversion with low effort. The probability of emission becomes much higher than that of absorption with "just a handful of electrons". Or in other words, if the only electrons of your system are in the upper laser level, even if just a few, the only probable thing is that they will de-excite and emit a photon.
This is of course simplifying things a bit, not going to extreme pumping or lasing, but should clear the confusion.
Decided to add another point:
What you describe above is also called the transparency point. At that point (lets call it 50/50) the net gain is 0. You sometimes absorb, sometimes emit. You need to picture all these processes with the energy level scheme though. You reach that point by pumping just a bit that the excitation gets enough electrons to match the number of electrons on the lower and not very populated level.