Do you have that level 17+ "railroad" (read: rollercoaster) in mind at level 5?
Yes (actually, from long before level 1 - I plan out the basic structure of a cmapaign years in advance - I have three 20 level campaigns outlined for future use right now - 2 revised versions of something I've run before and one new one) and no.
I have a primary plan, and I lay seeds for it. As the PCs go through the first 17 levels the pieces can become more clear - and usually the railroad at level 17 is triggered by an event that is inevitable and can't be stopped ...
Except players and PCs can sometimes come up with real good plans. Or, they just say, "Not my problem" and let a disaster that only they could stop happen. Either way, my 17+ railroad can end up falling apart.
So, if the players come up with a way to avoid the seemingly inevitable, I let them do it and then move on to a plan B. To be honest, this is very rare. When it does come up, I have tools built in - not tools specifically designed for this circumstance, but ones I can pull on to make it work.
For example, I borrow from future campaign setup. I lay groundwork in one campaign for the next campaign. I can use that future facing groundwork to insitute a backup plan. For example, in one campaign, one low level PC found a piece of the Rod of 7 Parts. The pieces become exponentially more powerful when you combine them, so one piece, by itself, is a good uncommon magic item. The original idea was that in a subsequent campaign, the PCs would spend a considerable amount of time hunting down the pieces and would run into the retired NPC (or their descendant) ... but if the PCs, at level 15, crash and wreck my plans for a level 17+ railroad in the campaign where one of the PCs has that piece of the Rod, I might instead have an NPC force gathering the rest of the pieces in the background and then have them come after the PC that had the one piece right around level 17 ... kicking off a different way to wrap up the campaign - but still built upon elements that were there throughout the campaign.
It is always a bit of an adjustment at higher levels as I tailor the game for the players and PCs at my table. My sandbox technique begins to thin the available options that I intentionally create for PCs as they get higher in level. This achieves a few goals, one of which is them feeling like they actually made a difference in the world. However, it does leave it so that they have a narrowed obvious set of DM provided options. I used to expect this to be a problem area where the players would feel squeezed after having so much freedom ...
But I find that by these levels, if the players are invested in your world, this is typically not a problem because the players will create more adventure hooks than they can close through their interest in something you mention casually. At times, the players that maintain a list of hooks can have hundreds of additional hooks they want to investigate that I just considered background lore ... but could be fun to build upon if they want to put their interest in that direction. The players tend to realize you were not necessarily 'building that for them', but I do make it clear that they can pursue anything they want in my setting.
