Optimizing PostgreSQL hstore for lots of small updates

Question

I have a situation where I will have one web session per row, but this session will generate dozens of updates to an hstore field in that row, one k/v at a time. At the end, I will have a completed structure. The keys will be more or less the same per session, but will evolve over time. The values will be mixed, with some being unique per session (e.g. email) and some having lots of repetition (e.g. male/female).

I am running under the assumption that lock contention will only be present at the row level, which is fine. What other gotchas should I look out for? I have zero experience with hstore, so I really want to make sure I understand what I am in for. Any feedback is appreciated.