If there are series of upgrades and you are late, you could burn from one to the next to the next, so there isn't any hindrance there.
You can only do this if the old coins even functions, isn't weakened and attacked, etc..
Actually you could have it in the protocol of the old coin that any coin it can be burned to (perhaps has to be registered in the old coin's block chain) can burn the coin for the old coin. But this requires a different sort of block chain consensus algorithm that doesn't assume it has all the data at every mining node all the time. This is close to describing my radical design for a block chain.
And if the old coin disrespects its own protocol, then one assumes the market will highly devalue it as dysfunctional.
That in turn will be reflected in the market as a sort of black swan risk that is extremely difficult to reason about or handle, leading to things being generally unstable. That is one reason why I find spin-off claims to be preferable. Once the claim date is passed, the claims are frozen, the state of the old coin no longer matters, and claims are not perishable.
But this has another risk which is the expiration on the demonstration of relative interest in the new coin, although that can also be signaled ongoing by relative market prices. Timing risks include a bug found in the new coin or some FUD campaign that needs time to be discredited that needs to be fixed first and people delay burning.
Yet I agree that while the period of burning is open, it complicates calculations of future value because the future coin supply is variable. It seems the best balance is probably some multi-month period, perhaps even up to a year.
Other risks during that longer period?
The possibility remains that people would prefer the old coin, since the only actual transition process is that the value of the old coin declines and the new coin increases (other than that, people are free to continue to use the old one if they prefer), but in a situation where ongoing development is occurring, presumably people would strongly prefer the new-and-improved version (or if not, is it really improved?)
It is complex though. The coin supply of the old coin is decreasing, so some might view it as gaining value, yet if mining declines too much some might view it as becoming more in danger of failing to function. To lower risks in burning, merge mining both the old and new is how I would code the upgrade which also facilitates the redesign of the block chain consensus I alluded to.
Claims can be made against a new coin with another ICO, but the added dilution adds to the risk that the market will reject it. Such is the nature of true decentralization. You can't simply corral people to the new coin if you don't deliver enough value. (This may also weaken confidence if there isn't a strong social contract, since people have no idea how much they will be diluted in future versions, nor how many times.)
I agree the additional unpredictable dilution causes discord in the social contract and confusion. Risk is a countervailing force to adoption.