[CI] test_vxlan_ecmp.py consistently failing on kvmtest-t1-lag-vpp

[rustiqly]

Description

vxlan/test_vxlan_ecmp.py is consistently failing on the kvmtest-t1-lag-vpp CI pipeline, causing the VPP test job to fail on virtually all PRs.

Evidence

Checked across multiple unrelated PRs (both rustiqly and others):

PR	Build	VPP Result	Failing Test
#22589	1047276	❌ FAIL	vxlan/test_vxlan_ecmp.py (2 failures)
#22591	1047540	❌ FAIL	vxlan/test_vxlan_ecmp.py (2 failures)
#22592	1047541	❌ FAIL	vxlan/test_vxlan_ecmp.py (2 failures)
#22594	1047367	❌ FAIL	vxlan/test_vxlan_ecmp.py (2 failures)
#22650	1047563	❌ FAIL	vxlan/test_vxlan_ecmp.py (2 failures)
#22581	1045657	❌ FAIL	Test plan stuck & timeout
#22575	1045455	✅ PASS	— (occasional pass)

The test fails with RUN_TEST_CASE_FAILED after 2 attempts, which triggers the test plan stop policy.

Impact

• The kvmtest-t1-lag-vpp job is marked [OPTIONAL] so it doesn't block merges
• However, the overall Azure.sonic-mgmt check shows as FAILURE, causing confusion for PR authors
• No PR code changes are related — this is a pre-existing/infrastructure issue

Topology & Test Details

• Pipeline: kvmtest-t1-lag-vpp
• Test file: tests/vxlan/test_vxlan_ecmp.py
• The test explicitly supports VPP (asic_type == 'vpp' handled in setUp fixture, line 188)
• No recent changes to tests/vxlan/test_vxlan_ecmp.py in master

Additional: kvmtest-t2 Failures

The kvmtest-t2 pipeline also frequently fails with two separate issues:

1. LOCK_TESTBED_FAILED: 7200s timeout waiting for testbed allocation (infra/resource contention)
2. everflow/test_everflow_testbed.py: Intermittent test failures on t2 topology
3. telemetry/test_telemetry_poll.py: Intermittent failures on t2

[bpar9]

this is a duplicate of #22662