Because the rules are usually unclear. Whether feedback is attributed, who sees it, and what it affects are decided during the cycle, and people answer differently depending on what they assume.
Structured feedback gathered from colleagues, peers and reports, rather than only from the reporting line.
Attribution and visibility are set when the cycle opens and shown to everyone giving feedback. Requests go to named people rather than an open call.
A cycle states that feedback is shown to the manager attributed and to the employee summarised. Everyone answers knowing that, which is the only way the answers are worth reading.
It does not anonymise robustly. In a team of four, attribution is obvious regardless of what the setting says — and pretending otherwise is worse than being clear.
Feedback attaches to the review cycle and the personnel record, visible per the rules set at the start.
It can be unattributed, but see the limits above — in small teams that is thin cover.
The manager or the employee, depending on how you configure the cycle.
With the review cycle, per your retention policy.
Fourteen days, every module, no card. Or half an hour with someone who will run it on your own records and tell you where it does not help.