I'm not sure what you want to achieve exactly but bloom filters are really trivial in how they work and thus easy to serialize. Now as you are talking about use across implementations, you would need some serialization standard or come up with one. I know there are libraries that allow serialization but apparently the serializations they consume or produce are not compatible from one library to the other.

If you have a compelling use case for a nip for example, I bet we can come up with a standard for nostr. Are you working on something?

With regards to your original question, I think it's easier and more efficient to maybe tag lists in queries so you can refer to these tags in subsequent queries as 67kB big data structures aren't great to keep around with every pubkey for example and neither are 67kB great to index.