Currently, if one defines report accelerations on a 'job server' isolated search-head, other search-heads will not be aware of the summaries created on the indexers. These summaries are scoped to the search-head or search-head pool where the report acceleration was defined, so unless they are part of the same search-head pool, the other search-heads will not be aware of their presence, nor is there are why to make them manually aware.
Note that segregating the scheduling of report acceleration on a job server would yield almost no benefit, as report acceleration activity happens only on the indexers and the search-head does almost no work other than record-keeping.
That being said, if you feel strongly that this functionality should be provided (for the purpose of type-based search activity segregation, for example), by all means please file an enhancement request for it.
