The key here is to examine the problems with GPS and the alternatives. These are the commonly cited problems:
1. It is possible to jam the signal,
2. It is possible for the set to fail without reporting it has failed,
3. It may be subject to accuracy degradation,
4. The database may be inaccurate.
Whilst there has been much talk of the ability to jam GPS signals and the availability of equipment to do so, there are no unscheduled reports of jamming. Whilst VFR GPS are not RAIM compliant if the signal is jammed they are far more likely to report a loss of signal or degradation of navigation to 2D.
It would be a poor pilot who was not monitoring his predicted en route course against the GPS by some other means. For example, with a moving map GPS you would have a pretty good idea how quickly the aircraft was moving along the map, the heading compared with that of your DI and your position by reference to visual features. In reality only the most subtle of the problems identified might go unnoticed. For example were the signal to be jammed the GPS would indicate a loss of signal and the aircraft would stop moving on the display or perhaps jump around the map in an unpredictable manner. Both are likely to get the pilots attention! A loss of degradation is far more subtle and could mean you are wide of your predicted course but it seems very unlikely that the degradation would be reflected in a continuos and predictable manner so that the aircraft was shown consistently 5 miles wide of track to the north for example.
In short a pilot who was at the very least cross checking his GPS with visual features occasionally and monitoring his progress to ensure it was following his predicted course is very unlikely not to be aware of any of these problems. That does of course mean he still needs to have the ability to fall back on some other form of navigation should the need arise.
The problems are broadly not dissimilar for a pilot en route operating IFR in IMC, other than the obvious issue that visual navigation is not now an option. As they have been for a long time now the tools of trade are the VOR’s and the PLOG. Once again if both are being monitored with even a modicum of diligence it is unlikely the discrepancy between the PLOG, the VORs and the GPS will go unnoticed for very long.
There is of course a significant difference with IMC operations (with broad similarities to flying very close to a zone boundary) and VMC operations in that during the landing and departure phases very accurate navigation is required and even for a pilot with good currency a risk that he may not identify a problem with the GPS of the sort described quickly enough. For this reason it seems vital that either operations of this sort are supported by other navigational aids or in the case of an approach the minimum takes account of the navigational risk. For example at an airport with an NDB DME procedure and a minima of 600 feet there would seem little problem relying on a RAIM GPS approach if the highest obstacle within the sector is 250 feet!
In short I cant help feeling a great deal of rubbish is printed about the use of GPS by people who have little understanding of its problems and its practical use. Of course if you are blindly relying on GPS for navigation and are an inexperienced pilot you may infringe a zone or end of getting hopelessly lost, but this must be weighted against the likelihood of the same pilot getting lost using traditional forms of navigation. In so far as the IFR pilot is concerned if he is going to operate in IMC in a SEP, SP, then he needs all the help he can get. He is hopefully going to have the experience not to rely on any one single means of navigation and that includes VORs. In over 1,000 hours of using a GPS in a SEP I have had the VOR fail twice, and become unstable or out of range on more occasions than I can count and yet the GPS has never (repeat never) failed. For such operations the aircraft is going to be fitted with at least one VOR, probably two, a DME and two moving map GPSs - there is pretty good redundancy and cross checking cover.