Selecting only a small amount of trials in a possibly huge condition file in a pseudo-randomized way

I am using the PsychoPy Builder and have used the code only rudimentary. Now I'm having a problem for which I think coding is inevitable, but I have no idea how to do it and so far, I didn't find helpful answers in the net.

I have an experiment with pictures of 3 valences (negative, neutral, positive). In one of the corners of the pictures, additional pictures (letters and numbers) can appear (randomly in one of the 4 positions) in random latencies.

All in all, with all combinations (taken the identity of the letters/numbers into account), I have more than 2000 trial possibilities. But I only need 72 trials, with the condition that each valence appears 24 times (or: each of the 36 pictures 2 times) and each latency 36 times. Thus, the valence and latency should be counterbalanced, but the positions and the identities of the letters and numbers can be random. However, in a specific rate, (in 25% of the trials) no letters/ numbers should apear in the corners.

Is there a way to do it?