Geometric dilution of precision is not an error as such, but an unfortunate consequence of the geometry of the satellites at a particular moment. It is better to think of this in two dimensions with two VORVOR —VHF Omnidirectional Ranges before extending it to GNSSGNSS —Global Navigation Satellite System and three dimensions.
If you have two VORVOR —VHF Omnidirectional Ranges and you are directly on a line joining them and between the two VORVOR —VHF Omnidirectional Ranges, then the two position lines you draw from the VORVOR —VHF Omnidirectional Ranges will overlap exactly. Because they overlap, there is no way of knowing whereabouts on that overlap you are. On the other hand, if you are not between, them, the position lines will not overlap - they will intersect, ideally at one point, with the "best" crossing of the lines being at 90°.
A similar sort of thing happens in three dimensions. If the position spheres from the satellites overlap, there is a degree of ambiguity about your actual position. Again, this is not because the system is somehow malfunctioning - merely a consequence of where you happen to be in relation to the satellites at the time.