Does a known-good camera work, powered by POE, when connected to the same port of the NVR? And how does the receptacle (and cable plug) for that camera's port on the NVR look? And what if you try both the original camera and a known-good (different) camera connected to that same port on the NVR, but using a different cable? And of course, you can try the suspect camera on a known-good NVR port using a known-good cable.
Gotta eliminate the camera's POE power converter, the NVR's POE power output, and the cable itself.
One of the nice things about Ethernet connections is that the signals are coupled through pulse transformers at both ends. This eliminates any direct DC connection between the Ethernet devices, and this affords some degree of surge protection as well as reducing noise issues by providing balanced, isolated connections. The breakdown voltage ratings for these coupling transformers vary from model to model, but can be several thousand volts.
On the other hand, the POE connections must have a direct DC path (be directly connected) from one end to the other. But the power converters at both ends, if designed well, will provide isolation, and if the PC board layout is done carefully on both ends, some protection is provided there, as well. Still, the converters themselves can be more vulnerable to these surge events than are the Ethernet ports.
Ideally, the POE outputs on the switch (or NVR) are all independent, using separate converters to maintain some isolation between channels. But this is up to the designer of the switch or NVR, of course. And the breakdown voltage between channels and between any channel and the main workings of the switch will depend on the design.
But you could potentially see damage to the POE converters on one or both ends from a surge event, and the cable itself can be wrecked by having its insulation damaged anywhere within the run. Of course, the Ethernet connections can also be damaged by surges, but in this case, we've already established that those wires within the cable must be at least workable, and the camera and NVR are working in that regard as well.
And if water did get into the connection, it could have fried the power converter in the NVR for that channel, but ideally, it would protect itself, limiting the current until the fault was gone. As
@wtimothyholman pointed out, the water could have been boiled away by the current passing through it until it dried enough to remove the conductive path.
If the DC-DC converter in the camera was damaged, but the rest of the camera operates, you could theoretically use that camera in a position where it would be practical to power it with the wall wart. Maybe not for a customer, but perhaps at home or around the office, etc., where you could tolerate future problems if it was compromised in some way.
Anyhow, the cable needs to be eliminated, and the POE power from that port of the NVR needs to be eliminated as well. Finally, the camera needs to be tested on a known good port of the NVR using a known-good cable to determine if it's really got a bad POE converter or not. This is almost quicker to test than it is to describe.
