About a decade ago I was in the same boat, until I played Paizo's We Be Goblins oneshot and we completely fucked up by starting the final fight in the worst possible position, leading to a TPK that made perfect sense with the story so far and had us falling out of our chairs laughing. With a little inspiration from a statblock I saw online I threw together a oneshot system that specifically intended to end each session by killing all the player characters, and ran it with my group several times to great success.

You consider a TPK a failure of GMing because to you, a TPK is a mistake - a misbalanced encounter or overly punishing consequences - that unintentionally and prematurely ends a story, but properly used it's a storytelling tool that intentionally progresses the plot and creates new events and options. In the context of my campaign, this was the second encounter and first of several combats with the final boss, that happen long before the party have any way to win the fight, nevermind permanently destroy her. She's such an extreme threat that PF2e's encounter building guidelines don't have a name for it. The encounter should have demonstrated her monstrous power by not only killing some or all of the party, but by killing them with the lingering effects of her spells after she had already left, but because I forgot the second save I completely failed at that. Every death in that fight would not only have reinforced her role as the BBEG after having been little more than a rumour over the last year of play, but also itself been the backstory of a whole new character - the mother who comes out of retirement to avenge her daughter's death, the aging priest who must undertake one final task for his god, the blacksmith that feels a sense of duty to those who once protected him.

Certainly TPKs can be a failure of GM, but like with many things it's entirely dependent on the context of your table, system, and story, and thinking it's always a failure of GMing misunderstands storytelling.