Perhaps the key (ha!, pun) is that you're talking about using RAM _and_ disk with the RAM being for caching/fast access that eventually hits the disk. Whereas, I think, in this case sqlite is either on the disk, or in RAM. There is no multiple tiers.
Correct me if I'm off here. Thanks.
This is closer to what I mean but not quite. Specifically, in-memory (or main-memory) is a technical term that talks about a specific type of database that focuses on doing operations /mostly/ in memory, with some spillover to disk as necessary, but only as an edge case. You usually then handle persistence with a sequential transaction log, perhaps in NVDIMM storage if you have some. This is in contrast to other systems where the memory is a buffer, but the "actual stuff" happens on disk. There is of course many hybrid schemes - it's really more of a spectrum. Some examples: SAP HANA, ArangoDB, MapD, MemSQL, MS SQL Server Hekaton.
There's also a lot of techniques that come with this style of thinking - often these are column stores, and often since these databases don't really hit disk, the bottleneck is moving data in and out of memory to the CPU caches, and sometimes just CPU speed. Often these use lightweight compression and SIMD instructions to tackle those problems.
So SQLite's "in-memory" database scheme doesn't quite count as something like this, it's more of a disposable database. But that's ok - it's not bad, it's just a different thing.
> Whereas, I think, in this case sqlite is either on the disk, or in RAM. There is no multiple tiers.
See the thread above - I kind of assumed this was the case, but like you mentioned, it looks like with the LSM implementation it does kind of sort of have a memory component - it lasts until you close the database "connection", and gets flushed often.