I guess this depends on how frequent it should be among drafters, since a theme could be "aggro" which is pretty general or "Equipment and aura boros" which is a little more specific. Ultimately this should end up in a persons deck, which means 23 playables usually. Not all 23 need to belong to the theme, in some cases mustn't if we're in a slightly more parasitic theme (i.e. not all decks want self mill cards), which means maybe half of the playables should be part of the theme?
If we want some variety to the theme then probably the drafter should have some on-theme cards left in their sideboard, so let's say half of their 45 picks are on-theme. That leaves us with about 22-23 cards out of 360, but maybe we want there to be some margin for error for the drafter to have on-theme cards snapped up by others, so we might bump that number higher. I think 30 seems like a nice round number? That'd be about 8% of a 360 card cube.
Pretty bogus estimates here, but I think working backwards from what would show up in the drafters 23 non-land cards is a good place to start. I would (as always) also recommend actually trying to build that 40 card test deck before hand just to get a feeling for what cards you'd actually slot in. Getting to play test and tweak it is ofc also good, then you'd have some data about what cards you probably need in the deck, so that you can then say "if I need X such cards and what probabilities Y and Z I will make these design choices".