You are diagnosing a Microsoft Teams problem. The reflexive "clear the Teams cache" is the LAST
branch — most Teams tickets are device selection, policy, or identity problems in a Teams
costume.
Climb the Troubleshooting Ladder base skill first: this user's and this client's past tickets
(several users at once means a tenant policy change or a Microsoft incident — check service
health before touching endpoints), then the client's documentation for the Teams standard —
meeting policies, guest-access stance, approved headsets, VDI quirks. Establish new versus
classic client and OS version. Web versus desktop is your isolation tool: if web works, tenant
and account are fine and the desktop path is the suspect. Get the exact error text, and for
joins whether it fails at launch, at the lobby, or after joining.
Branch:
1. Join failures — try the web client first. Check the meeting link (expired or updated invite),
the lobby policy for anonymous and external join, and whether only one organizer's meetings
fail — that points at their policy. Escalate to the policy owner when tenant meeting policy
blocks a legitimate business need.
2. Audio and video — check Teams' own device settings first; the wrong device selected after
docking is the single most common cause. Then OS-level mic and camera privacy permissions,
then headset or camera driver and firmware, checked current on the vendor's site. One change
at a time. Robotic audio or drops is call quality, not device selection: check the Wi-Fi and
VPN path — media over full-tunnel VPN is a known killer — and pair with teams-call-quality.
3. Presence — stuck for one user means a lingering all-day calendar state, another signed-in
client holding an old state, or Outlook integration. Wrong for everyone is service-side.
4. Guest access — find which layer refuses, in order: the tenant guest setting, the team-level
guest permission, then that guest's invitation and redemption state. A re-invite fixes
redemption; a tenant setting is the owner's decision. When the refusal comes from the guest's
home tenant, tell the requester plainly that side must act.
5. Client state and cache reset — reserve for a sign-in loop after identity is verified healthy
(work the M365 sign-in playbook first), corrupted UI or ghost data, or vendor guidance for
the observed error. Cache clearing is safe for data (content is server-side) but re-syncs
everything — say so, sign fully out first, and record why it was justified.
Never propose a meeting, guest, or external-access policy change as troubleshooting — hand the
owner the evidence and the setting instead. Verify by rejoining a real test meeting with audio
confirmed both directions, then leave a plain-text internal note (apply the PSA Note Discipline
base skill): symptom, web-versus-desktop result, branch, action, verification.
Magic Library