Should I jump to 3.10.7 now or wait for 4.0?

On your color question: nothing changes unless you opt in. The new color correction is only reachable through the Channels API, so if your sketch never calls it, the compiler just skips that whole chunk of code and your strip keeps doing exactly what it does now. There's no default-on behavior that would shift your colors under you. That was actually the whole point of tying it to the Channels API, so existing sketches wouldn't silently carry the extra code or the extra behavior.

On the size stuff, your read is roughly right. The middle releases added a bunch of code that got linked into every sketch wheather you used it or not, and 3.10.7 walks most of that back. From what I've seen on my own builds, the smaller boards are back under where 3.10.3 was, and the ESP32 family is close to even, maybe a touch bigger. The slight bump there comes from the driver rework that watches for network activity and adjusts buffers to keep output stable, wich I'd count as a fair trade.

Should you move with a Halloween deadline? My honest take: if 3.10.3 is working and you're flashing a finished build for an event, leave it alone. The classic advice of don't change versions days before a deadline applies. The update isn't going to break your colors, but any version jump carries some risk of a compile hiccup or a driver quirk on your specific board, and you don't need that stress right now.

That said, I did make the jump on two ESP32 boards a couple weeks back and it was boring in the best way. Flashed, compiled, ran, done. So it's not a scary release. If you get through Halloween and want the smaller binaries, update then and give yourself an afternoon to test. Waiting for that last bugfix build before 4.0 is also a perfectly fine plan, since it'll be the most polished point in this line either way.

Appreciate the detail here, though my situation is a bit diffrent since I'm still on an old Uno driving a short strip. Size limits aren't my problem, the opposite if anything. My worry is more long term, like if I stay on 3.10.3 forever do I eventually hit a wall where new example sketches just don't compile anymore? I updated mid-project once before and spent a weekend chasing a weird flicker I never did fully explain, so I've been nervous about touching versions since. Is there usually a decent gap between a release like this and 4.0 actually landing, or do these things drop faster than you'd expect?

Been on the same FastLED version for years on my holiday displays. Learned my lesson back when I used to chase every release, spent one December swapping library files the night before a party and the whole thing refused to compile til sunrise. Now I keep a folder with the old library zip sitting next to every project, so if anything ever goes sideways I can roll back in two minutes. That way updating is low stakes, you try it after the deadline and you always have a parachute. One more thing, test your fallback folder actually works before you need it, a backup you never flashed once is just a hope.

I jumped to 3.10.7 a few weeks back on my porch build and it just worked. My trick with any library change is I flash the new version on a spare strip and cheap board first, let it run a couple days, then move the real project over. Costs me nothing and it caught a bad PSU pairing once that woud've been a nightmare to debug inside the actual enclosure. One thing I noticed after updating was my compile time actually got shorter, not something I expected but I'll take it. If you've got a spare strip lying around you could do the same and be fully confident before Halloween.

The thing that changed all this for me was moving my projects to PlatformIO. I pin the exact FastLED version in each project's config file, so updating the library on my machine never touches a working build unless I deliberately edit one line. Before that I was on Arduino IDE and every library update hit every sketch at once, which is how I burned a Saturday re-flashing stuff that was fine the week before. With the deadline you've got, staying on 3.10.3 is the right call, but if you're still on the IDE, this might be the nudge to try pinning. Makes the whole update question a lot less stressful.