The water below the ice remains liquid since geothermal heating balances the heat loss at the ice surface. The pressure causes the melting point of water to be below 0ºC. The ceiling of the subglacial lake will be at the level where the pressure melting point of water intersects the temperature gradient. In Lake Vostok the ice over the lake is thus much thicker than the ice sheet around it.
The water in the lake can have a floating level much above the level of the ground threshold. In fact, theoretically a sub-glacial lake can even exist on the top of a hill, provided that the ice over it is so much thinner that it creates the required hydrostatic seal.
The floating level can be thought of as the water level in a hole drilled through the ice into the lake. It is equivalent to the level that a piece of the ice over it would float at, if it was a normal ice shelf. The ceiling can therefore be conceptualized as an ice shelf that is grounded along its entire perimeter, why it has been called a captured ice shelf. As it moves over the lake it enters the lake at the floating line, and leaves it at the grounding line.
For the lake to exist there must be a hydrostatic seal along the entire perimeter, if the floating level is higher than the threshold. A hydrostatic seal is created when the ice is so much higher around the lake that the equipotential surface dips down into impermeable ground. Water from underneath this ice rim is then pressed back into the lake by the hydrostatic seal. The ice surface is ten times more important than the bed surface in creating the hydrostatic seal. This means that a 1 m rise in the ice surface at the ice rim is as efficient as a 10 m rise in the bed level below it. In Lake Vostok the ice rim has been estimated to a mere 7 m, while the floating level is about 3 km above the lake ceiling (Erlingsson 2006).
If the hydrostatic seal is penetrated when the floating level is high, the water will start flowing out in a jökulhlaup. Due to melting of the channel the discharge increases exponentially, unless other processes allow the discharge to increase even faster. Due to the high head that can be achieved in subglacial lakes, jökulhlaups may reach very high rates of discharge.
There is also evidence that there are subglacial lakes on Jupiter's moon Europa. Not all lakes with perennial ice cover can be called sub-glacial, though, since there are also those that are covered by regular lake ice. A criterion for glacial ice is that it is flowing. Ice needs to be ca 30 m thick to start flowing, so frozen over lakes are not likely ever to turn into subglacial lakes by themselves.