Thought I’d shed some light on this with my own experience using grooms.
In my opinion, grooms are not suitable for realtime gameplay in most circumstances yet. There are exceptions to this- The Windwalker Echo project uses grooms for example, and seems to maintain a playable framerate on my machine (RTX 2060 Super). However, the hair isn’t especially detailed and there is only one character on screen.
If you have a really powerful machine, it might be fine running a few grooms. But if you intend to distribute your game, keep in mind that according to steam hardware surveys as of Sep 2022 the most common GPU is a GTX 1060, which can barely handle a single groom, if at all. So while YOU might be able to run a bunch of grooms just fine, others will not.
There are some things you can do to improve groom performance. Reduce strand count, decimate the curves/vertices, use LODs, etc. And while these will provide some performance gain, grooms are so expensive to begin with that the gain simply isn’t enough to justify usage most of the time. It is absolutely not worth it to have your FPS drop from 120 to 60 or less for a single groom hairstyle.
I would imagine that you can get even better results by using Maya xgen or similar and creating hairstyles specifically meant for realtime gameplay (ie, very low strand count, tweaked over multiple iterations for performant results). But creating realtime hair this way is almost an entire career path in itself, and certainly too much for an indie dev to concern themselves with.
If you are creating a game that only has a single character that uses groom hair, you can probably get away with it with some optimization tricks. But if you intend to have multiple characters on screen, all with their own groom hair, save yourself the frustration and use hair cards instead. Banging your head against the wall figuring out ways to get grooms to be more performant is a waste of time, and hair cards can look quite good with decent materials.