Person creation failure prolonged by cache

[tiina303]

Metabase query:

select error_timestamp, error, count(*) as cnt from events_dead_letter_queue where error_timestamp > toDateTime('2022-06-20 00:00:00', 'UTC') and error_timestamp < toDateTime('2022-06-23 00:00:00', 'UTC') and error LIKE '%1818c218a10c6-050c126d844092-4c356833-59b90-1818c218a1217%' group by error_timestamp, error order by cnt desc

We can see that events from June 22, 2022, 3:56 PM until June 22, 2022, 4:32 PM go to DLQ for this person.

What happened here is we cached the person here, which updates the cache so later we know it's not a new person,
but due to the Postgres being in read-only mode around 15:56 UTC the creation failed.

The cache duration is 4h, but likely we saw error only 30min due to the user logging off/ending the session.

If we didn't cache it, then as soon as the db was allowing writes again we'd be creating the person and getting the rest of the info for their session.

Related issue: #10396

[tiina303]

this has been resolved