Well, usually everything depends on the job requirements and the reliability of an algorithm. Databases don't usually recreate ids based on whether you delete a record or not for a number of reasons:
1 - the counter auto-increments in a way that reliably generates UNIQUE ids;
2 - it's totally agnostic of any holes created by deleted records;
3 - which makes it simple to prevent duplicated ids
As an added bonus the process mimics the accounting numbering analogical systems in which accountants just invalidated previous issued documents but the next id was going to be the next of the sequence of already attributed ids, hence totally predictable.
Whether you are using one method or the other, just make sure you can rely on the system to generate unique ids, and don't accidentally reset the generator, which would eventually cause you troubles in the long run.
Another suggestion would be to generate some sort of more complex alphanumeric serial id (totally independent of the form request id or the item_id) with a timestamp inserted on it. I use these all the time to make sure that not only the IDs are unique but also that they are temporally traceable and still in a sequence that could be considered incremental, although not composed of consecutive +1 serials.
If needed, I'd still use an extra field in LISE to insert the form request id if you need to pair the submission to the LISE item.
In any case these are just suggestions, just go along with what works best for you and your requirements.