How can embedded platforms generate IVs but face trouble generating nonces? Aren't IVs also nonces? Or do you mean a different aspect of it is the problem?
Typically with a re-used IV the same plain text will lead to the same cipher text, which can be an issue.
On the other hand with a re-used nonce you can basically recover the xor of two plain texts from two cipher texts. This leads to plain text very quickly and is a much more serious issue.
Since most embedded platforms have limited entropy and/or no hardware RNG it is more difficult to generate proper distinctive, unpredictable iv's (in the case of CBC). If most of the nonce is random and the rest predictable (like GCM) the same problem applies. The nonce for GCM for instance is at least 8 bytes random in most implementations.
There are enough ways to generate sufficiently randomized iv's. People just don't bother usually. (guess 0 as an iv for instance on many embedded systems).