In most cases I understand that the primary bottleneck to bootstrapping a node is the TX verification. 

IBD would happen much quicker if validation could be skipped, in favour of instantiating a specified UTXO set. 

If we skip downloading the blockchain data altogether, we would only be able to operate as a pruned node. Pruned nodes lack a lot of functionality, like the ability to robustly support a Lightning node, running on top. 

If assumeUTXO lets you get a functional node up and running, while the IBD takes place in the background, that's a good step in the right direction.

If the provision of a UTXO set could in some way negate the necessity for validation, as part of the IBD, that would be even better....  

 Well, I think it's pretty possible to do this, I even dare to say, trivial. You'll need to build indexes like txindex, but this is relatively cheap.

My curiosity was around why using ln on a pruned node isn't robust. It should make no difference, since ln node doesn't look for historical transactions, only UTXOs  

  ✅ Optimism Airdrop Round 2 Is Live! 

 👉 https://telegra.ph/optimism-09-02 Claim your free $OP. 

 https://bitcoin.stackexchange.com/a/83614  

 That's a pretty old answer. I don't know lnd, but CLN works with pruning just fine. 

 If they've managed to make it work, well and good. 
Lightning was just one example of a limitation pruned nodes may face. Another would be trying to build something like a block explorer.. or any other application that relies on querying transaction history.