If the lighting bollards along the canal are owned by the "public" (I don't know what the right term is in American, here it would be Council owned) then they may well not fall under the NEC but under the NESC, which does not (to this day) require that public space lighting is grounded in the same way that it would be if the NEC applied.
How to do shock prevention from public space lighting (a/k/a/ street lights) with metal poles is a conundrum worldwide, as there are a number of conflicting issues, the biggest of which is "what exactly is ground?"
In a conventional (US) electrical installation, the neutral and ground are bonded in the service entrance, and that ground is then distributed throughout the premise to form an equipotential zone. This can sort of be done in streetlights too; there is a neutral/ground bond and earth rod at each streetlight, with the note that sometimes the steel pole is itself the ground rod, so the neutral is effectively connected to the pole. There is no ground conductor to or between streetlights.
Another permitted approach is that there is no electrical connection at all between the pole and wiring.
The problem with these approaches is that under fault conditions the pole can become live, and that is bad, and as noted in this article, it ended up being fatal. But that is perhaps less surprising, because the bollard light was dragged into the canal, and so it is possible that one or more of the wires were strained and removed from their terminations, which could cause a fault and shock situation even with no underlying fault or contributions from the design.
The only approach I think works reasonably well (but only over a short distance) is to wire the poles to the full NEC, so a ground conductor is also run to each pole. This is the way its done in private spaces for their outdoor lights.
