We either roll in front of each other - in the first session, or for online games we use an online roller. So there's no real way to cheat other than using loaded dice.
Sure there is; suicide the character you don't like by doing stupidly dangerous things. Repeat until you get a result you like. Takes longer, sure, but you'll get what you're looking for eventually... you'll just annoy the rest of the group if you keep doing it too much.
Which is why the notion of Point Buy and Arrays are so popular in modern systems. It saves a lot of time and frustration by just letting the players start out with a PC they actually want to use while also preventing other forms of cheating (because the GM can easily do a "check sum" of the scores and they don't have to break from the rest of the PCs to watch them roll in the middle of a session because their last character died to make sure they aren't massaging the results).
I know random stats is a huge part of the OSR experience... and that's just one of many reasons I'm NOT an OSR fan. In my three and a half decades of experience the only thing random stat rolls are good for is stories of funny PC deaths as those who rolled crap run through PCs in ridiculous ways until they get one that doesn't suck (and the occasional even funnier story of a crap PC that a player keeps trying to suicide and fails because the dice keep saving them).
Personally, I prefer to skip that step and just get on with an actual campaign where we're playing real characters and not collections of stats we haven't even named yet because we're not sure they'll survive their first session. If I wanted THAT I'd play a more in-depth board game; same tactical decisions, no pretenses that they're actual characters with drives and goals making in-universe decisions instead of cardboard cutouts you're waiting to see if its worth even slapping a name onto.
Hell, EVEN Palladium Books; godfather of never changing core mechanics; has acknowledged how nonsense completely random stats are in RPGs. Their second edition Robotech RPG (the one with The Shadow Chronicles as part of it) actually lets you choose one of eight arrays (based on which stat you want to be highest) where the variance is, for example;
Fast Reflexes and High Dexterity: I.Q. 1D4+10, M.E. 1D6+9, M.A. 1D6+8, P.S. 1D6+9, P.P. 1D6+19, P.E. 1D6+9, P.B. 1D6+10, Spd 1D6+17.
So the absolute WORST you can possibly do with that is; IQ 11, ME 10, MA 9, PS 10, PP 20 (+3 to strike/parry/dodge), PE 10, PB 11, Spd 18.
The absolute best is (remember 15 or less is no bonus); IQ 14, ME 15, MA 14, PS 15, PP 25 (+5 to strike/parry/dodge), PE 15, PB 16 (30% charm/impress), Spd 23.
Basically, the clamp on best/worst is now +2 to strike/parry/dodge (or about +/-5% from the average result) if you decide on that build.
If even Palladium Books is abandoning completely random rolls for an almost pre-determined result (it's Palladium, they can only bend so far), that just about says it all.