Searching within very large subset of documents

I have a medium-sized (~10m) Lucene index and I frequently want to =
repeatedly search within a subset of around ~100k documents. I can =
increase MaxClauseCount and build up a huge TermQuery, keep that around, =
then build a BooleanQuery out of the result at runtime, but the =
resulting query is quite slow. The now deprecated Filter would have been =
a good option with a BitSet, but that=E2=80=99s deprecated.

Any thoughts on the best way to do this?

Thanks!
-twb