Even after initial sync? It shouldn't keep many dirty buffers after IBD because data only gets added once every 10 minutes on average.
Discussion
I had many issues with this in Floresta, because I didn't want to flush too often (this takes up to multiple Ms). But if I didn't, this could put the system In a undefined state.
My goal was to allow recovering from unclean shutdown, in older state. This roughly what core does
Same case as discussed in https://github.com/cculianu/Fulcrum/issues/41?
I’d say it’s not much of a hassle considering Floresta is meant to be compact. Telling that the solution is to resync Fulcrum is a bit too much. It takes roughly 12 hours on a modern laptop with plenty of memory and I/O and it takes 2 weeks with a raspberry pi class device.
For a single user it is dawning to me that the tradeoffs for the amazing speeds Fulcrum offers are not that great and maybe smaller implementations like Electrs or soon Florets are a more sane choice.
If this happens again I’m switching back to electrs.
Yeah, in Floresta I have the advantage of storing old states an starting from there in case of failure.
For users, I honestly don't get why do a normal user needs a full indexer. It's super redundant and creates a lot of trouble for users.