feat: add resilient reconnect and catch-up replay

README.md

@@ -10,6 +10,10 @@ Current implementation targets the normal SuiteLink tag protocol and supports:
 - subscribe and unadvise flows
 - update decoding for `bool`, `int32`, `float32`, and `string`
 - write (`POKE`) encoding for `bool`, `int32`, `float32`, and `string`
+- background receive loop for subscription updates without manual polling
+- automatic reconnect with durable subscription replay after runtime disconnects
+- best-effort latest-value catch-up replay after reconnect when enabled
+- policy-based reconnect retry timing with exponential backoff and jitter support
 - client/session/transport layers suitable for macOS, Linux, and Windows
 
 ## Unsupported
@@ -18,8 +22,9 @@ This repository does not currently support:
 
 - AlarmMgr / alarms and events
 - secure SuiteLink V3 / TLS transport
-- automatic reconnect
-- background receive loop / production retry behavior
+- write queuing while reconnecting
+- full outage-history replay of every missed value
+- validated reconnect behavior against a live AVEVA server deployment
 - validated support for richer System Platform data types such as `double`, `int64`, or `DateTime`
 
 ## Build
@@ -56,3 +61,9 @@ See [README.md](/Users/dohertj2/Desktop/suitelinkclient/tests/SuiteLink.Client.I
 
 - The repository includes fixture-backed protocol tests under [Fixtures](/Users/dohertj2/Desktop/suitelinkclient/tests/SuiteLink.Client.Tests/Fixtures).
 - Protocol assumptions derived from reverse engineering are intentionally isolated in codec classes and tests so they can be refined against live captures later.
+- Normal subscription use does not require calling `ProcessIncomingAsync`; the runtime loop dispatches updates in the background.
+- After a runtime disconnect, the client enters `Reconnecting`, rebuilds the transport/session startup sequence, replays durable subscriptions, and resumes dispatch without caller resubscription.
+- Reconnect timing is driven by `SuiteLinkRetryPolicy`; the default policy retries immediately, then uses bounded exponential backoff with jitter.
+- Catch-up replay is best-effort latest-value refresh only. It does not represent every value missed during the outage window.
+- If the client is reconnecting after a runtime disconnect, `WriteAsync` fails fast until the session is ready again.
+- Reconnect behavior is currently verified with fixture-backed and fake-transport tests; live AVEVA validation is still required, especially for mixed-mode deployments and server-specific timing.