(Untested.)
Suppose you have a layer of aluminum foil or gold leaf on top of a piece of glass and you vaporize a pinhole in the metal with an arc. You can position the electrode that creates the arc to submicron precision, and by controlling the energy of the arc, you can vaporize a precisely controlled amount of the metal, creating a hole comparable to or slightly smaller than the thickness of the foil — household aluminum foil is about 10 μm, aluminized-mylar Doritos bags are about 1 μm, and gold leaf is about 0.1 μm. Smaller controlled thicknesses may be feasible if deposited onto a substrate like boPET or polyimide, for example by vacuum coating.
Some of the vaporized metal will impact the glass, blowing air out of the way in the process, and condense there, thus locally depositing metal on the glass.
This provides a simple way to locally deposit a fairly precisely controlled amount of metal onto a substrate, insulating or conductive, with submicron positional control. Aside from the potential utility of marking the surface of the glass or other substrate, by repeating the process layer by layer, you can manufacture arbitrary three-dimensional shapes from the metal with submicron precision, and much higher material deposition rates than are possible with electroforming.
Doing this with noble metals such as gold, platinum, or iridium should be possible even in air, but more reactive metals such as silver, copper, aluminum, and iron probably require an inert-gas atmosphere, or at least nitrogen. Conductive oxides like those of lead or silver might permit the use of this process for simultaneous arc deposition and reduction in a reducing atmosphere such as acetylene, but hydrogen contamination of the resulting metal might reduce the utility of this approach.
Electric arcs can easily reach temperatures high enough to sublimate even carbon, so this process can deposit even very refractory metals like tungsten, molybdenum, or tantalum. Moreover, by putting a little distance between the substrate and the feedstock, very thin films can be formed.
In most cases, though, minimizing the distance between the substrate and the feedstock would be desirable. One way to achieve this is to cut channels at known locations into the surface of the substrate, say a one-millimeter grid of 100-micron-wide channels, and pull a light vacuum on those channels. In that way the gas necessary to sustain the arc in the process as described above can be employed to eliminate the unwanted gap rather than sustaining it.
Under vacuum, instead of an arc, a high-power electron beam or focused ion beam could instead be used to locally vaporize the feedstock, as in e-beam etching and FIB milling. This should permit nanometer resolution rather than the mere submicron resolution routinely attainable with mechanically positioned tooltips.
Of course, the localized heating of the glass substrate may also have effects, desirable or otherwise. If they are desirable, the optimum “feedstock” may be a film of conducting graphene or graphite rather than a metal. And, as Russ of Sarbar Multimedia points out, you can use a laser rather than an arc to heat the feedstock, and if the substrate is transparent, you can place the feedstock on the opposite side of the substrate so that the laser side of the feedstock is the one touching the substrate.