I see in many situations, people believe there are limits of how many times an event in roulette can happen, or not happen.
I do not think there are such limit, but at the edge the probability for it should happen is very low, and needs a very large sample.
I did for fun a script which sounds crazy at the first view, but if we should find a method hardly ever lose, it must go bust very very rare, and still make some plus. This is mathematical not possible on a balanced wheel or proper RNG.
If we remove the zero (and there are NoZero games) or use table with zero but have no house advantage, it should be possible to run a progression, which test the "limits", and not is a "break even" play until it lose.
I have run 64000 spins and until then it has not lost a session (It can any trial). It use the spread which exist on some tables with one cent min. We play straight ups and use a negative progression.
This I have tried real just a few times, and won easy, as nothing strange happen, a hit within 70 spins.
During simulation of 64000 spins my highest bet was 63 units, and then a payout of 36 times that.
We could run up to a bet of 5000 or more units and be inside the table limit.
If you try the simulator, and find a win on bet zero, it is the first bet winning, as the count starts with zero.
It tells the pay out and there is a net win every hit, how much we reduce the bets done from the pay out.
friwebb.se/rulle/pleinmart/ is the link.