issue list: counts are period-wide, not scoped to the release filter passed in --query #1583

Summary

sentry issue list --query 'release:"<release>"' correctly restricts which groups are returned, but the count and userCount in the response are period-wide totals across every release, not the counts matching the query. Nothing in the output signals this, so a release-scoped triage built on issue list silently over-reports by an arbitrary factor.

Environment

sentry 0.40.0
node v24.18.0
Windows 10.0.26200

Steps to reproduce

Any project with several releases contributing to the same issue group.

# 1. list issues restricted to one release
sentry issue list <org>/<project> --period 14d \
  --query 'release:"<release-A>" is:unresolved' \
  --sort freq --limit 100 --fresh --json

# 2. ask the events endpoint for the same group, with the same release filter
sentry api "/organizations/<org>/events/?field=count()&field=count_unique(user)\
&query=issue%3A<SHORT-ID>%20release%3A%22<release-A>%22\
&statsPeriod=14d&project=<project-id>&dataset=errors" --json

Actual

For one group in a project with 21 releases:

The issue list numbers equal that group's totals over the period across all releases. The factor is not constant — it depends on how the group is distributed across releases, so it cannot be corrected after the fact.

Expected

Either the counts reflect the applied query filter, or the output marks them as unfiltered lifetime/period totals so a caller does not attribute them to the filter.

Impact

This is the documented path for release triage (references/skill guidance suggests issue list --query 'release:"X" is:unresolved' --sort freq), and it is the natural entry point for agents. In my case it promoted an unrelated high-volume group to "top problem in release X" when that group's actual contribution to the release was a small fraction of the reported number — the release only had ~8k events total while the tool reported a single group at 4118 for it.

Notes

• --sort freq ordering appears to follow the same unfiltered counts, so the ranking is affected too, not just the displayed numbers.
• The events endpoint returns correctly scoped values, so the data is available; it looks like the issues endpoint's stats simply are not filter-aware.
• Related but distinct: sentry issue list --json --fields omits count, firstSeen, lastSeen, userCount (missing expand=stats) cli#969 (fields omitted without expand=stats) — that one is about which fields appear, this one is about the values being out of scope.