There are reasons to undervolt a card (optimising performance and efficiency, by squeezing as much performance while making it use less lower, generate less heat, etc)

But undervolting it to fix a bad connector design isn't it.

Most problems stem from cables not being fully seated. If you have a direct 12v2x6 (I hate this name), that reduces failure points, as long as make absolute sure that the cable is fully inserted.

Older PSUs that don't have that connector, use the PSU manufactuer supplied (or a good quality third party cable if you have those for aesthetics) direct 2/3pin PCIe to 12v2x6 cable.

The adapters that are supplied with the GPU are "fine" but you should avoid using them if you can because they increase potential failure points.

The issue stems that a GPU will still function even if the cables aren't entirely fully inserted, but because the connection isn't clean, and because they don't load balance appropriately or just refuse to work if a full connection isn't made, it will draw power down terminals outside of their specs. That much current generates heat, too much heat and it melts.

They've tried to fix this by making the terminals shorter (12V2X6 is essentially 12VHPWR but with shorter pins), but it's a bad implementation of a way to try and simplify a connector and board traces.

At least with old PCIe connectors, if they weren't making a connection the GPU would refuse to run so it forced you to recheck your cabling.

Most cases are user error but it's a user error introduced with a bad design, because there aren't any safeguards to prevent it at all. This has happened to even experienced builders.

Make sure absolutely everything is in as far as it will go.