There’s a mathematical relation between the size of a bubble (actually, the radius of curvature) and the pressure inside the bubble: basically, the smaller the radius of curvature, the higher the pressure.
This is why scratches are nucleation sites:
a bubble forming inside a small scratch has a much greater radius of curvature for the gas/liquid interface than a spherical bubble of the same volume would have, thereby decreasing the pressure necessary to form the bubble.
In your hypothetical case where there are no good nucleation sites, eventully a bubble will form on a “bad” one: turbulence, a shock wave caused by you walking past the flask, or even just a momentary defect in the liquid itself caused by the random motion of molecules.
Once a bubble does form, a large fraction of the liquid will decide that this is the chance it has been waiting for (it’s much easier to increase the size of an existing bubble than to form one in the first place) and join the party, whereupon the bubble grows very quickly and splashes the rest of the liquid all over the lab (not that I would have ever done anything like this myself; someone, uh, told me about it, yeah, that’s it…). This is called “bumping”.
To avoid this, it’s common practice to add boiling chips (generally small pieces of inert porous material; sometimes glass beads are used instead) to liquids before heating them. This provides lots of nice nucleation sites and avoids bumping, provided of course that you remember to put in the boiling chips
first and don’t try to add them when the liquid’s already at the boiling point.