Pausing cURL to Clear Output and Resuming - bash

Question

I hope this is not off-topic here. This is the background:
My friend and I are trying to solve this online puzzle where you basically have to enter a 16 digit combination; if wrong, the site'll display "incorrect" and, if right, some other message we don't know about. We have the first eight numbers so we figured it would be feasible attempting to brute-force it using curl and saving to output to check.
We've implemented it like this:
curl http://www.game.com/checkCombo.php?num1=[0-9]&num2=[0-9]...&num8=[0-9] >> output.txt
As you can see, we're basically running through all possible combinations and appending each individual output to the "output.txt" file.
The way we check whether we've gotten the right combination, since we don't know what it'll say then, is by searching the number of "incorrects"; if when we open the file we've gone through 1000 combinations and there are 1000 matches for "incorrect" then we know we haven't hit the right one yet. If when we open the file we've gone through 2000 combos and there are only 1999 matches for "incorrect", we know we've hit the right combo and we just search for it.
However, since we're saving 10^8=100,000,000 individual outputs, we're running out of space on the computer.
We tried pausing the cURL process with "ctrl+z" and checking the file to see whether we've hit the right combo. If not, we clear the "output.txt" file, save it, and resume the cURL process with "fg %#".
The problem is that when we resume cURL, after we've cleared the text file, the file remains empty: the output of cURL won't save any more onto the file.
I'd like to know whether there's a way of pausing cURL to check the file, clearing the file, and resuming curl while continuing to save the remaining outputs.
If you have other ideas on how we might go about the combination, I'd also appreciate hearing about them. Thank you.

Answer 1

A simplified case will show different things to get started. Start testing without curl until you have a working solution.
I have written a funtction checkcombo that needs 3 parameters and has a hidden solution:
checkcombo () {
digit=$1$2$3
if [[ $digit = 412 ]]; then
echo "How did you find me?"
else
echo incorrect
fi
}
Try this function with
checkcombo 4 1 0
checkcombo 4 1 1
checkcombo 4 1 2
Nice, now we want bruteforce. Perhaps the best solution is one loop until 999 and split each number in 3 digits, but I want to show the nested loops.
for i in {0..9}; do
for i2 in {0..9}; do
for i3 in {0..9}; do
checkcombo $i $i2 $i3 | grep -q "incorrect" || { echo The answer is $i $i2 $i3 && break 3; }
done
done
done
The line with checkcombo can be hard to read. It is a short way of saying
if [ $(checkcombo $i $i2 $i3 | grep "incorrect" | wc -l) -eq 0 ]; then
echo "The answer is $i $i2 $i3"
# and stop testing with break 3
fi