BadTimeout (0x800A0000) — The operation timed out

The server did not respond within the client's timeout. It says nothing about whether the server eventually processed the request, which matters for writes.

Start free OPC UA → canonical

Diagnosis

When you hit this

Reading hundreds of nodes in one request against a server with a small per-request budget.
A subscription publishing interval faster than the server can actually sample.
Network latency on a routed or VPN link where the client kept an SDK default tuned for a LAN.

Resolution

How to fix it

Batch smaller
Split large reads into chunks. Servers commonly advertise MaxNodesPerRead — honour it rather than discovering it by timeout.
Match publishing interval to reality
Asking for 100 ms from a server sampling at 1 s produces timeouts and no extra data.
Treat write timeouts as unknown, not failed
A timed-out write may have landed. Read the value back before retrying, or you risk applying it twice.

Normalization

What Forge does about it

A request that fails on some points and succeeds on others still normalizes. The tags that arrived resolve to canonical fields as usual; the ones that did not come back are listed in unresolved_tags and excluded from coverage_pct, so a partial read reports as partial instead of quietly looking complete.