[mysql] Activity job crashes with TypeError when event_timer_end is NULL

[MatheusSandre]

Agent Environment

Datadog Agent 7 (public.ecr.aws/datadog/agent:7), running on ECS Fargate.
MySQL 8.0 (AWS RDS), DBM enabled.

Describe what happened

The query-activity job of the MySQL DBM check crashes on every run:

File "/opt/datadog-agent/embedded/lib/python3.13/site-packages/datadog_checks/mysql/activity.py", line 308, in _normalize_rows
    if row["event_timer_end"] < seen[row["thread_id"]]["event_timer_start"]:
TypeError: '<' not supported between instances of 'NoneType' and 'int'

Followed by:

(utils.py:453) | [... job:query-activity] Job loop crash

_normalize_rows compares event_timer_end against a previously seen event_timer_start without a NULL guard. In performance_schema.events_statements_current, TIMER_END is NULL while a statement is still executing, so any activity sample that catches an in-flight statement crashes the job.

The crash repeats on every collection interval. Regular mysql.* metrics keep flowing; only query activity/samples stop being collected.

Describe what you expected

Rows with a NULL event_timer_end should be skipped or treated as still-running, not crash the job loop.

Steps to reproduce the issue

1. Enable DBM on a MySQL instance with query_activity enabled.
2. Have a statement running long enough to be sampled while still in flight (TIMER_END NULL in events_statements_current).
3. The activity job crashes with the traceback above.

Frequency: reproduces consistently on a moderately loaded instance — observed on every collection cycle over several hours.