I have done something somewhat similar-- and in 4e, too!-- so I have at least some relevant insight. But, be warned, I am a thoroughly urban person with no experience of rivers except to drive over them on bridges, so my advice is not very river-oriented.
You've Identified The Problem
You've correctly identified the problem, I think, especially in comparing your corridor dungeon to this situation: Straight corridors, like linear rivers, don't offer much in the way of choices. They can, and probably will, degenerate into railroad-like slogs, just one darn encounter after another without much the players can do about it.
A Frame Challenge: Embrace Extended Skill Challenges
What you should consider doing, here, is making the river itself into an extended skill challenge. What I did once that was similar was model a race from Point A (where a creature was encountered in a forest) to Point B (a location in a clearing that the players already knew of, and knew the creature was heading for to cause mischief.)
I think this was very similar to your case because they didn't have a lot of choice-- they didn't know the forest well enough to plan some alternate route, they were left mainly following in its wake trying to catch up. So conceptually, for that session, they were on geographical rails.
So I modeled this as the biggest Skill Challenge that 4e would support-- 4 out of 7, 5 out of 9, whatever it was-- and came up with an appropriate number of forest-themed mini-chases, with as much of a spread of relevant skills as I could manage. So there was a swift stream that can to be crossed on slick stepping stones (definite challenge to the less dextrous), recognizing the signs of a swarm of hornets that the creature disturbed on its way past (a choice of taking damage, vs losing ground), a section where the path ran off a minor sheer drop (skill checks to see if they get down safely, or a choice of easier skill checks and time loss to rig ropes up, etc.)
The idea behind all of these is that the players/characters could analyze a small situation, propose some courses of action that involved rolling dice and if they did well they lost nothing; if they did poorly, they would typically lose health, some resources, or time. (Remember, this was a chase.)
This was a pure chef's kiss of a session-- everyone was invested, everyone was throwing their skills at the challenges, and they were feeling the time pressure because I would report whether they gained or lost ground.
If you can re-frame your river adventure like this, I think there is a real chance of success. (This does not have to be a chase, by the way, it just be a looming deadline. Or it could be a chase where they are the pursued party, inverting my model.)
If you're intent on keeping combats as part of this, I still think this model can help at least somewhat. But now the idea is that failing those challenges results in combat while succeeding does not. Or, succeeding means an easier fight, or a fight on better terms or terrain. Etc.
The Keys To This Technique
There are two keys to this technique:
Player choice and interactivity. What this technique adds back in to the linear railroad is player choice and engagement. It is still A - B - C - D - E. But at each node in that linear graph, the characters have freedom and agency and the chance to shape it. This is crucial.
The players MUST KNOW in some fashion the results of their successes and failures. If they saved or lost time in a race, you have to tell them that, otherwise they will not know. If they avoided a combat, or made a combat easier (or harder) you have to narrate that, otherwise they will not know.