The risks from gain-of-function mutations, at least on whole viruses, are pretty large compared with the benefits, at least in the short term. Let’s say you find a bat virus, induce a mutation, and then find that the mutation makes it highly infectious to humans. What’s the plan, then? Vaccinate everyone against that? Well, that’s pretty damn expensive, and unless you’re going to share with literally everyone on the planet, and not just, say, your own country, you’re going to get accusations of designing a bioweapon. Are you going to do that for every functional mutation you find? That’s going to add up quickly, and you can pump out new mutations faster than we can verify the safety and efficacy of a new vaccine and then pump out 7 billion doses. Essentially anything you’d design, whether vaccine or small molecule drug, would take considerable expense to produce and would be largely worthless unless it actually did turn into a pandemic, and if you actually start a pandemic with a new mutation, then the prior vaccines and drugs may be ineffective and again a waste of effort.
There’s potentially some good work that can be done on gain-of-function mutants of isolated proteins from the virus, but you don’t need to have viruses that can infect cells in order to do that. You can have bacteria or yeast produce those proteins en masse for you from templates that contain only the code for just those proteins on plasmids. For example, you could do random mutagenesis on just the spike protein and find sequences that don’t bind human antibodies generated by the immune response to one or more of the vaccines and crosscheck those for the continued ability to bind ACE2. That could potentially alert you to some sequences that would be cause for extra alarm should they pop up in viral surveillance sequencing, and there’s no risk of starting a pandemic with it (outside of someone literally and maliciously manipulating whole viruses according to your data to (re-)start a pandemic, I suppose).