bereng

This is a POC diag PR so far. Yes this will get propagated once complete.

On this one you'd need a CNDB PR imo to make sure we don't break anything there?

Dtests seem broken on jenkins atm. CI failures seem to align to the last successful run used as baseline. The extra Map operations should be a drop in the ocean...

> A side note: won't we exhibit the same kind of problem with I left this one alone as: A. `invalidateAll()` lies in an unused code path atm B. We'd...

@pkolaczk a quick review for a second opinion here would be appreciated if you have a gap.

^I'll ping you when were both online see how you connected the dots as I didn't see that!

So we talked about this and cleared the smoke a bit: - Blocks I noticed such as 0x000000058b679aa8 are blocked waiting on nothing. We are going to assume that is...

The last commit rollsback the deadlock fix I had introduced and adds precomputed prepared statements sizes. If we like it we can send it see if it passes the deadlock...

> we shifted it from doing N units of work every Nth insertion to doing a unit of work (computation) on every insertion Previously we were doing 2 times the...

Rebased and squashed. @pkolaczk would you be so kind to complete your review please? I intend to merge this today/tomorrow. Thx.