Flaky `test_rest_headers` suite

[cthulhu-rider]

• ubuntu (fail) https://rest.fs.neo.org/HXSaMJXk2g8C14ht8HSi7BBaiYZ1HeWh2xnWPGQCg4H6/3450-1750691261/index.html
• macos (pass) https://rest.fs.neo.org/HXSaMJXk2g8C14ht8HSi7BBaiYZ1HeWh2xnWPGQCg4H6/3450-1750691127/index.html
• ubuntu (pass) https://rest.fs.neo.org/HXSaMJXk2g8C14ht8HSi7BBaiYZ1HeWh2xnWPGQCg4H6/3450-1750702289/index.html

Current Behavior

fail occured once

Expected Behavior

either always pass or fail

Possible Solution

perhaps the problem is on NeoFS side. Need to explore

i also noticed that in failure case same OID occured in different suites. In such cases, a read-delete race is usually suspected

Context

nspcc-dev/neofs-node#3406

Regression

maybe

[cthulhu-rider]

not reproduced locally after 5 attempts

[evgeniiz321]

@roman-khimov seems like a bug in rest gw for me, there are 2 objects with same attribute - 'Writer': 'Leo Tolstoy'. Test removes one, then tries to get a second object by this attribute but gets an error status: code = 2052 message = object already removed. get by id works ok. Not sure why it reproduced once, maybe some caching issue?

[roman-khimov]

It looks like cache or search returning an old ID which is then unavailable. Depending on specifics it's either a feature or a bug. @evgeniiz321, can you look into logs to check what really happens there?

[evgeniiz321]

waiting for a reproduction