(Speaking with DM hat on...)
I prefer they stick to a baseline, but I deal with it.
My main issue with power creep isn't the power creep itself, but just the fact that it "stretches out" power-level expectations among gamers. If there's a well-established baseline, everyone's typically got a similar power level assumption in mind. Otoh, if the game's power level grows a lot over time, then some gamers might come to the table aiming at the original power level while others want the current higher power level.
No biggie really, but it's just one extra expecation to manage in session zero.
While blatant power creep should not be a desired outcome, a little creep is inevitable, and irrelevant so long as it is not excessive. So long as players can run their PCs to high level without the PCs becoming 'outdated' by power creep, there is no real problem.
From a pragmatic perspective: Creating a game system means coming up with dozens of interlocking bits that play off each other and create new and inventive ways to achieve your goals. No designer, no matter how prescient or skilled, can foresee every possible interaction and, more than that, foresee every possible way to create new material for the game.
And as you create that new material you'll run into interactions that increase the level of power in the game. Sometimes because the new material is too flexible or too powerful, but often because it combines with other materials either produced or new that create unintentional interactions...
And, of course, that's presuming you got the "Balance" right in the first place. Later passes over the same material may show that your initial designs don't line up with your actual goals after they've slammed into public play.
So power creep isn't that huge a deal, for me, in the end, so long as the baseline gets pulled up to deal with it, or the outlying issues get appropriately nerfed.
If unnecessary? Usually yes. If for an option that has been underperforming powerwise since their publication? No, it's usually a good thing, so long as the "creep" is specifically measured to not make the option overpowered. 5e has plenty of examples of this, from Dragonborn to Rangers to Sorcerers, and so on.
The answer, as usual, is "it's complicated". Power creep is a tool, and like all tools, it can be used in good ways. But it can also be used in bad ways.
Some amount of power creep is inevitable for any game that continues to expand content. If well-managed, it is perfectly fine, but that’s not a small if. It should be treated as a necessary evil, managed and kept to as gradual a rate as possible.
I think 5e has managed its power creep quite well overall.