As someone who has been through this exact thing, it sucks, but it is possible in the current court systems. Courts have complicated "Discovery" processes explicitly for this reason: to find out not just the immediate problem (a bankrupt builder or a builder who would be bankrupt by the end of the court case), but the supply lines (what other warranties apply?) and the creditors (where is the bankrupt company's money going?).

A human run court is allowed to state "this bankruptcy looks fishy for this reason, and these people are owed damages, and some of the money that moved on already to these creditors is by rights should be paid for damages to these people". It's definitely not fun to be involved in such court proceedings, and it is and likely always be an inefficient human process, but it is a process as a society we've built to be mostly very trustworthy. A jury of peers to decide if damages are warranted in the first place; a human judge with reasoning to say "this shell company is bankrupt, but these other companies can and should pay, maybe with the money they took from this shell company" or "this shell company is bankrupt but some of this still applies to the warranty insurance product they purchased and the insurance provider needs to pay up even though their client is bankrupt".

It doesn't always work right/decide in your favor, and it is slow and expensive, but "smart contracts" offer no better and often a lot worse processes. A "smart contract" can't route around a shell company/wallet. A "smart contract" likely can't discover that other warranties may have applied or other insurance products existed (and if it can it's in a programming class all its own and likely a one off you can't expect to exist in general or to work if you need to actually rely on it).