This is not an engineering issue, it is a navigation issue.
Military-grade inertial navigation systems (like an ESGN) have been meters-over-a-month accurate for at least 40 years. In practical terms, a vessel so-equipped would not need an external fix (determination of position) over an entire voyage, even one months long. That, however, does not remove the requirement for obtaining external fixes. Failure to take external fixes is what led to the USS Atlanta (SSN-712) verifying that fixed objects (like Gibraltar) are stronger than even nuclear submarines. They thought the specifications of their inertial navigation was sufficient; it was not.
Submarines must use their inertial systems and dead reckoning to determine their position while submerged. These are vessels traveling at 10-25+ knots while essentially blindfolded, because sonar only keeps you from hitting things that make noise. Navigation standards do not allow using solely internal sources to set the vessel's position. There are too many ways errors can creep into any internal system, and the result of being wrong can be catostrophic. So the ship's inertial position is periodically reset via external systems, such as GPS. GPS is extremely accurate, but is not the only method available, Satellite Navigation (SatNav) being the other primary method (uses doppler calculations from moving satelites). These will be used to reset the ship's inertial systems to the latest fix. In the absence of electronic aids, celestial navigation is also available, along with terrestrial (visual) fixes if close to land.
So no matter how good commercially available inertial navigation systems are, a source for external fixes is always required.
Source: my US Navy officer's qualification in submarines (gold dolphins).