- update the watcher to merge consecutive categorized child build ids into one logical grouped subrun when they resolve to the same real host-based group
- prevent one real grouped run from being fragmented across multiple watcher subruns just because the raw ci-build-id label changes between hosts
- keep grouped Mattermost posting aligned with the inferred group identity instead of the unstable raw child build id
- update the categorized watcher to use the latest matching host reporter artifact written before grouped finalization so grouped results can be posted as soon as the host run is actually done
- reduce dependence on the later parent Cloud Run Finished summary block for grouped-run completion reporting
- keep grouped runs from waiting until near parent-run completion when real host-level completion evidence is already available
- keep categorized grouped sub-runs in RUNNING state when only the grouped check-xml-files.ts artifact has appeared and the parent log has not yet exposed the real host summary
- prevent blank Mattermost grouped-run posts with empty host tables, zero counts, and zero/invalid timing caused by posting too early
- require the categorized watcher to wait for host-level grouped summary data before treating the grouped run as complete
- update the watcher to stop trusting misleading categorized child build labels when they do not match the host/spec actually being executed
- infer the reported categorized group name from the actual host being run, so mismatched labels like ubuntu-batch for a Red Hat host are corrected in status reporting
- document the categorized watcher workaround in the ATVM guide, watcher design, and watcher README without changing the underlying ATVM runner scripts
- relax grouped-run duration parsing so Linux categorized summaries like `12m 19.4` and `15m 42.2` are converted reliably even when the trailing `s` is split awkwardly in the parent run log
- keep the categorized grouped-run summary extraction aligned with the host results already being parsed from the parent Cloud Run Finished blocks
- refresh the controller watcher copy so the next categorized run uses the improved grouped duration parser
- update the watcher to parse completed categorized grouped-run host summaries from the parent run log instead of relying only on grouped XML files that often contain only check-xml-files.ts
- add grouped-run duration parsing so categorized sub-run timing can be derived from the Cloud Run Finished summary when host XML details are absent
- fix the completed-summary to grouped-xml alignment so filtered-out older artifacts do not shift host-summary assignment for the current run
- update the watcher to treat categorized parent-run activity as the authoritative signal for whether the overall request is still running
- prevent the watcher from exiting early just because one categorized grouped sub-run completed and wrote artifacts
- document that categorized watcher instances must remain alive between grouped runs until the parent request has actually gone inactive past the grace window
- update the ATVM guide, watcher design, and install docs to reflect the stricter categorized parent-run completion rule
- interpret ATVM controller log timestamps in the controller's local timezone before converting to UTC so the watcher uses the correct current-run window
- prevent newly started categorized runs from immediately picking up older categorized artifacts just because the parent build name was reused
- keep the categorized watcher focused on artifacts from the current controller run instead of stale prior attempts
- update the watcher to detect the active categorized sub-run from the live `--ci-build-id` process state instead of treating the parent run as one synthetic grouped run
- fix host XML parsing so the watcher prefers the real host suite over the `Root Suite` entry, avoiding `0 tests, 0 failures` summaries
- use the first timestamp inside the run log as the watcher start time so restarted watchers do not miss current-run categorized artifacts because of log file mtime drift
- improve active-host inference for categorized runs so the watcher maps the current categorized build to the correct host family while the sub-run is still in progress
- update the watcher start helper to stop any stale watcher instance for the same requested parent build name and remove its old state directory before starting fresh
- document that reused parent build names must not inherit stale cancelled, posted, state.json, or subruns state from older runs
- update the watcher install and design docs so the controller workflow explicitly treats stale reused-build-name state as part of startup cleanup
- update the watcher design and automation guide to treat --categorize as sequential ATVM sub-runs rather than one parent run with internal phases
- document that categorized runs should send one Mattermost status per completed grouped sub-run instead of one parent-only final post
- add a --categorize option to the watcher start helper so categorized mode is explicit in watcher startup
- update the watcher implementation to track categorized sub-runs separately, write per-subrun state, and post each completed grouped run once
- update the watcher cancel helper so it writes a final CANCELLED state into state.json before stopping the service
- record cancellation timestamps and a cancellation note in the watcher state file for clearer post-run inspection
- update the watcher service docs so the documented cancel behavior matches the state-file handling
- remove the committed watcher-service __pycache__ bytecode file from git tracking
- ignore Python bytecode artifacts so generated .pyc files do not get committed again
- keep the watcher-service source files as the only tracked implementation artifacts
- add the per-run ATVM watcher service package under atvm/watcher-service, including the Python watcher, systemd template unit, helper scripts, and deployment docs
- document the watcher-service install and operating model, including one-run-per-instance behavior, Mattermost posting rules, and the best-practice /opt/atvm-watcher-service install path
- clarify ATVM run approval semantics so `approve` means run without watcher and `approve with watcher` means run and start the watcher
- update the ATVM automation guide and AGENTS rules so watcher usage and approval behavior are explicit and consistent
