Actually no-one has replicated ball lightning reliably under experimental conditions. The topic is outside my field of expertise, though (or, for that matter, anybody's), so who knows? Assuming the personal accounts are accurate, it's unlikely that anyone will be able to replicate this in an experiment because:
- the mechanism is unknown;
- the conditions in which it emerges are unknown;
This is superficially similar to dowsing, I suppose, but there is a key difference: no-one (or at least, no large group of people) has ever claimed that they can make ball lightning occur in a simple way, whereas dowsing apparently just needs a pair of hands and two sticks, or a bent piece of metal, or similar. In that sense it is easier to test, and then accept or reject, dowsing than it is ball lightning. Additionally there exist plausible mechanisms for ball lightning, but to date no explanation for dowsing is even plausible, let alone accepted.
The ease of the test is important -- and then the results are very clear. Dowsing does not stand up (so far, at least) to scrutiny. Ball lightning I don't know about, but it's still very much in the realm of being poorly understood.