The format that is used during generation is totally unsuitable for probing of the tablebase, as it stores al winning distances separately, in huge chunks that can only be read sequentially. The data of all cycles thus has to be collected and merged into one format, which is then compressed in small stretches, so that quasi-random access for probing the winning-distance is possible.
The leapfrog generator does none of this!
A separate program will be needed to convert the raw tablebase in generation format to a format suitable for probing. And it is very unlikely that I will write it: I am not interested in random probing of tablbases, and only generate them for the purpose of collecting statistics on the various end-games.