diff --git a/CHANGELOG.md b/CHANGELOG.md
index 3462f03..4e0fa6f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -10,6 +10,12 @@ release. The format is loosely based on
 
 ## irtt-proto
 
+### 0.5.2
+
+#### Added
+
+- Exported `VERSION`, the compiled crate's Cargo package version, for applications reporting their component versions. Wire behavior is unchanged.
+
 ### 0.5.1
 
 #### Added
@@ -34,6 +40,25 @@ release. The format is loosely based on
 
 ## irtt-client
 
+### 0.7.0
+
+#### Added
+
+- Opt-in automatic session reconnection in the Tokio managed driver and `BlockingManagedClient`. Continuous measurement runs can retry initial connection failures, failed or peer-closed sessions, and server-limited session completion. Silence detection reconnects a session that stops replying while probes are being sent.
+- Each desired target retries independently, leaving healthy targets running. Exponential backoff doubles from `reconnect_initial_delay` (default 250 ms) to `reconnect_max_delay` (default 10 s); a valid reply restores the initial delay. Reopening resolves the original endpoint again. Target removal and explicit shutdown cancel recovery.
+- Exported `VERSION` for compiled component-version reporting.
+
+#### Fixed
+
+- Stopping or replacing a target during session opening cancels the pending open immediately, instead of waiting for its retry schedule to finish.
+
+#### Compatibility
+
+- `ManagedClientConfig` gains `reconnect_timeout`, `reconnect_initial_delay`, and `reconnect_max_delay`; update complete struct literals or use `..Default::default()`. Reconnection is disabled by default (`reconnect_timeout: None`) and is valid only for continuous normal measurement mode, with a positive schedulable silence timeout. Retry delays must be positive and schedulable, and the maximum must be at least twice the initial delay.
+- `ManagedConfigError` gains `InvalidReconnectPolicy` and `InvalidReconnectDelays`; downstream exhaustive matches must handle them. Delay validation applies even when recovery is disabled.
+- A reconnecting target returns to pending/opening status rather than producing a terminal outcome for each retry. Consumers must handle repeated session-start events for the same target instance and reset session-bound statistics and sequence tracking on each new session.
+- Use `irtt-stats` 0.6 with this client version when exchanging `ClientEvent` values. Tokio remains optional; the default low-level client remains runtime-free. No wire-protocol migration is required.
+
 ### 0.6.0
 
 #### Added
@@ -105,6 +130,12 @@ release. The format is loosely based on
 
 ## irtt-server
 
+### 0.5.4
+
+#### Added
+
+- Exported `VERSION`, the compiled crate's Cargo package version, for applications reporting their component versions. Server policy and wire behavior are unchanged.
+
 ### 0.5.3
 
 #### Added
@@ -150,6 +181,23 @@ release. The format is loosely based on
 
 ## irtt-stats
 
+### 0.6.0
+
+#### Changed
+
+- Updated the public `irtt-client` dependency to 0.7. Applications passing `ClientEvent` values into this crate must upgrade both dependencies together; client 0.6 and 0.7 event types come from different crate versions and cannot be mixed.
+- Added signed nanosecond changes to `IpdvPairUpdate`: `rtt_ipdv_ns`, `send_ipdv_ns`, and `receive_ipdv_ns`. Existing `Duration` fields and aggregate IPDV statistics remain absolute magnitudes. Update complete struct literals to include the new fields.
+- Exported `VERSION` for compiled component-version reporting.
+
+#### Fixed
+
+- Directional IPDV is available with midpoint timestamps, using the midpoint for both server endpoints.
+- Per-packet IPDV preserves the sign of the delay change instead of reporting only its magnitude.
+
+#### Compatibility
+
+- Collectors remain scoped to one target session. Reset or replace a collector when the managed client reconnects; this release does not aggregate across session sequence resets.
+
 ### 0.5.2
 
 #### Changed
@@ -192,6 +240,30 @@ release. The format is loosely based on
 
 ## irtt-rs
 
+### 0.10.0
+
+#### Added
+
+- Continuous CLI recovery with `--continuous --reconnect` (or `--duration 0 --reconnect`). Failed initial opens, failed or closed sessions, server-limited completion, and reply silence trigger independent per-target retries. Healthy targets keep running, including in multi-target and `--targets-stdin` runs; removing a desired target cancels its retries.
+- `--reconnect-timeout` configures reply-silence detection while sending (default 10 s). `--reconnect-initial-delay` and `--reconnect-max-delay` configure per-target exponential backoff (defaults 250 ms and 10 s); the maximum must be at least twice the initial delay. Valid replies reset the backoff. Explicit stops do not retry.
+- `--no-stats` disables client statistics collection, retention warnings, and final summaries while retaining measurement events and structured output columns. Statistics-dependent fields have absent values.
+- Optional `stats` Cargo feature, enabled by default and required by `tui`. Build a lightweight client without statistics using `cargo build --release -p irtt-rs --bin irtt-client --no-default-features --features client`; use `--features client,stats` for a client-only build with statistics.
+- `--continuous` as an explicit alias for zero duration, `-d` for duration, `-i` for interval, `-l` for packet length, client `-f` for output format, and `-v` for verbose output. Client/TUI durations accept fractional and composite values such as `2.5s` and `1m30s`.
+- Client/TUI `--timeouts` configures the session-opening attempt schedule independently of session reconnection.
+- Client `-n` / `--no-test` negotiates parameters and exits without measurement probes, using the selected output format.
+- `--version` on the dispatcher and applets; applets also report their compiled library versions and whether client statistics are disabled.
+
+#### Fixed
+
+- Signed per-packet IPDV is preserved in output; directional IPDV is available with midpoint timestamps.
+
+#### Compatibility
+
+- Reconnection is opt-in for the CLI and requires continuous measurement mode; it cannot be combined with `--no-test`. Statistics reset on every reopened session, so summaries describe only the latest session per target, not the whole continuous run. A warning explains this when recovery and statistics are both enabled. Continuous TUI recovery remains enabled as before.
+- Statistics are no longer implied by the `client` feature. Existing custom builds using `--no-default-features --features client` must add `stats` to retain collection and summaries. `--no-stats` is accepted in builds without statistics; output schemas are preserved, with absent IPDV values.
+- Nonzero duration values require units; zero (including `0s`) still selects continuous mode. `--continuous` and an explicit `--duration` are mutually exclusive.
+- Application-library users must account for the added client argument/configuration fields and optional statistics feature. Dependencies are now client 0.7 and stats 0.6, with proto >=0.5.2 and server >=0.5.4 for component-version reporting. MSRV remains Rust 1.88.
+
 ### 0.9.0
 
 #### Added
@@ -330,10 +402,15 @@ Starting with this release, crates are versioned and tagged independently
 rather than in lockstep. Each crate's release is tagged as `<crate>/vX.Y.Z`
 (e.g. `irtt-proto/v0.5.0`), pointing at the commit its version was released
 from. `irtt-proto`, `irtt-client`, `irtt-server`, and `irtt-stats` are
-libraries published to [crates.io](https://crates.io); their tags exist to
+libraries published to [crates.io](https://crates.io); publication is a separate
+manual Cargo step. Their tags exist to
 identify the exact source for a given crates.io release and do not produce a
 GitHub Release. The application package (`irtt-cli` through 0.5.0, `irtt-rs`
 from 0.6.0 onward) is the only package with prebuilt binary artifacts: pushing
 an `irtt-rs/vX.Y.Z` tag triggers [cargo-dist](https://opensource.axo.dev/cargo-dist/)
-to build and publish a GitHub Release with platform binaries. The historical
+to build and publish a GitHub Release with platform binaries. On Gitea, the
+same application tag triggers the server Docker image workflow, publishing
+`X.Y.Z` and `latest` to Gitea and GHCR. Library tags
+trigger the Gitea Semver workflow, but neither binary nor image publication.
+The application tag also triggers that Semver check. The historical
 `irtt-cli/v0.5.0` tag remains as released; it is not rewritten or replaced.
diff --git a/Cargo.lock b/Cargo.lock
index 14be465..d02e759 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -857,7 +857,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-client"
-version = "0.6.0"
+version = "0.7.0"
 dependencies = [
  "irtt-proto",
  "irtt-server",
@@ -879,7 +879,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-proto"
-version = "0.5.1"
+version = "0.5.2"
 dependencies = [
  "hmac",
  "md-5",
@@ -890,7 +890,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-rs"
-version = "0.9.0"
+version = "0.10.0"
 dependencies = [
  "chrono",
  "clap",
@@ -911,7 +911,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-server"
-version = "0.5.3"
+version = "0.5.4"
 dependencies = [
  "getrandom 0.4.3",
  "irtt-proto",
@@ -923,7 +923,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-stats"
-version = "0.5.2"
+version = "0.6.0"
 dependencies = [
  "irtt-client",
  "measurement-stats",
diff --git a/Cargo.toml b/Cargo.toml
index d66bc81..e466658 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -16,10 +16,10 @@ rust-version = "1.88"
 repository = "https://github.com/Lochnair/irtt-rs"
 
 [workspace.dependencies]
-irtt-proto = { version = "0.5.0", path = "crates/irtt-proto" }
-irtt-client = { version = "0.6.0", path = "crates/irtt-client" }
-irtt-server = { version = "0.5.2", path = "crates/irtt-server" }
-irtt-stats = { version = "0.5.2", path = "crates/irtt-stats" }
+irtt-proto = { version = "0.5.2", path = "crates/irtt-proto" }
+irtt-client = { version = "0.7.0", path = "crates/irtt-client" }
+irtt-server = { version = "0.5.4", path = "crates/irtt-server" }
+irtt-stats = { version = "0.6.0", path = "crates/irtt-stats" }
 
 [profile.release]
 opt-level = "z"
diff --git a/README.md b/README.md
index 85a631b..d14d3f8 100644
--- a/README.md
+++ b/README.md
@@ -45,8 +45,8 @@ this project's interop harness.
 | --- | --- | --- |
 | Wire protocol | Protocol version 1, verified against upstream v0.9.1 | Protocol version 1 (unchanged in v0.9.2) |
 | Finite probing | Supported, default 10s | Supported, default 1m |
-| Continuous/indefinite operation | Supported (`--duration 0`); still computes running statistics (see below) | Supported via streaming mode (`-s`, added in v0.9.2); computes no statistics at all in that mode |
-| Statistics in continuous mode | Running count/min/max/mean/variance per target; exact medians unavailable; bounded IPDV tracking | None in streaming mode — raw/event output only, by design, to avoid unbounded memory |
+| Continuous/indefinite operation | Supported (`--continuous` or `--duration 0`), with optional per-target reconnection | Supported via streaming mode (`-s`, added in v0.9.2); computes no statistics at all in that mode |
+| Statistics in continuous mode | Optional running count/min/max/mean/variance per session; no exact medians; bounded IPDV tracking. Use `--no-stats` or omit the `stats` feature to disable collection | None in streaming mode — raw/event output only, by design, to avoid unbounded memory |
 | Multi-target probing (one invocation) | Supported, with labels and staggered/burst pacing | Not supported — one remote address per invocation |
 | Terminal UI | `irtt-tui`, a live graph/dashboard applet | Not present |
 | Output formats | table, csv, tsv, jsonl, with selectable columns | Text summary, JSON (optionally gzip to file), and a fixed-field raw mode; no CSV/TSV and no column selection |
@@ -115,9 +115,16 @@ irtt-client --no-test netperf-eu.bufferbloat.net:2112
 Probe continuously until interrupted (`Ctrl-C`):
 
 ```sh
-irtt-client netperf-eu.bufferbloat.net:2112 --duration 0
+irtt-client netperf-eu.bufferbloat.net:2112 --continuous
+
+# Recover failed, closed, or silent sessions independently per target
+irtt-client netperf-eu.bufferbloat.net:2112 --continuous --reconnect \
+  --reconnect-timeout 10s --reconnect-initial-delay 250ms --reconnect-max-delay 10s
 ```
 
+Reconnection resets statistics for that target; summaries cover its latest
+session. Add `--no-stats` for measurement events without statistics collection.
+
 Watch the same session live in the terminal UI:
 
 ```sh
diff --git a/crates/irtt-app/Cargo.toml b/crates/irtt-app/Cargo.toml
index b13459b..4199a20 100644
--- a/crates/irtt-app/Cargo.toml
+++ b/crates/irtt-app/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "irtt-rs"
-version = "0.9.0"
+version = "0.10.0"
 description = "IRTT-compatible client, TUI, and server applications"
 edition.workspace = true
 license.workspace = true
diff --git a/crates/irtt-client/Cargo.toml b/crates/irtt-client/Cargo.toml
index eb66fb6..c7d22d1 100644
--- a/crates/irtt-client/Cargo.toml
+++ b/crates/irtt-client/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "irtt-client"
-version = "0.6.0"
+version = "0.7.0"
 description = "IRTT-compatible Rust client library"
 edition.workspace = true
 license.workspace = true
diff --git a/crates/irtt-client/README.md b/crates/irtt-client/README.md
index 9c59ef4..b9b053f 100644
--- a/crates/irtt-client/README.md
+++ b/crates/irtt-client/README.md
@@ -35,6 +35,21 @@ Use `ClientEvent::from(receipt)` for an `EchoSent` event with `scheduled_at` and
 `timer_error` absent; the managed driver supplies those schedule measurements
 and publishes committed sends even when post-send processing fails.
 
+Managed continuous measurement runs can opt into automatic recovery with
+`ManagedClientConfig.reconnect_timeout: Some(timeout)`. Initial connection
+failures, closed/failed sessions, server-limited completion, and reply silence
+while sending trigger independent per-target retries. `reconnect_initial_delay`
+(default 250 ms) doubles to `reconnect_max_delay` (default 10 s); a valid reply
+resets the delay. Both delays must be positive and schedulable, with the maximum
+at least twice the initial delay. Explicit stop and target removal cancel retries.
+Recovery is disabled by default and is invalid for finite or negotiation-only runs.
+
+In 0.7, complete `ManagedClientConfig` literals must include the three recovery
+fields (or use `..Default::default()`), and exhaustive `ManagedConfigError`
+matches must handle the new recovery-policy variants. A target can emit repeated
+session-start events within one instance; reset session-bound state, including
+statistics, for each reopened session. Use `irtt-stats` 0.6 with client 0.7 events.
+
 Tokio stays optional: the default build has no runtime and no Tokio
 dependency at all.
 
diff --git a/crates/irtt-proto/Cargo.toml b/crates/irtt-proto/Cargo.toml
index 7189157..4aed92c 100644
--- a/crates/irtt-proto/Cargo.toml
+++ b/crates/irtt-proto/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "irtt-proto"
-version = "0.5.1"
+version = "0.5.2"
 description = "IRTT protocol encoding and decoding primitives"
 edition.workspace = true
 license.workspace = true
diff --git a/crates/irtt-server/Cargo.toml b/crates/irtt-server/Cargo.toml
index bcb8f7b..20ba773 100644
--- a/crates/irtt-server/Cargo.toml
+++ b/crates/irtt-server/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "irtt-server"
-version = "0.5.3"
+version = "0.5.4"
 description = "IRTT-compatible Rust server library"
 edition.workspace = true
 license.workspace = true
diff --git a/crates/irtt-stats/Cargo.toml b/crates/irtt-stats/Cargo.toml
index 8be4096..b6388d3 100644
--- a/crates/irtt-stats/Cargo.toml
+++ b/crates/irtt-stats/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "irtt-stats"
-version = "0.5.2"
+version = "0.6.0"
 description = "Statistics aggregation for irtt-client events"
 edition.workspace = true
 license.workspace = true
diff --git a/crates/irtt-stats/README.md b/crates/irtt-stats/README.md
index b6e3709..71896f6 100644
--- a/crates/irtt-stats/README.md
+++ b/crates/irtt-stats/README.md
@@ -16,10 +16,16 @@ re-export. Event interpretation, IPDV pairing, loss, and snapshot reconstruction
 remain here.
 
 A `StatsCollector` has one sequence/IPDV namespace and one packet count, both
-scoped to a single target. A managed session covering multiple targets needs
-one collector per target — as the `irtt-rs` CLI does — since a target's
+scoped to a single target session. A managed session covering multiple targets
+needs one collector per target — as the `irtt-rs` CLI does — since a target's
 sequence numbers start over at zero and feeding two targets into one
-collector would pair up unrelated probes.
+collector would pair up unrelated probes. Reset or replace the collector when a
+target reconnects, since the new session restarts its sequence numbers.
+
+Version 0.6 consumes `irtt-client` 0.7 events. Applications using both crates
+must upgrade them together to avoid incompatible `ClientEvent` types.
+`IpdvPairUpdate` exposes signed nanosecond changes alongside the existing
+absolute `Duration` values; aggregate IPDV statistics remain absolute magnitudes.
 
 ## Retention modes
 
diff --git a/docs/irtt-client.md b/docs/irtt-client.md
index 5f483c9..d5719c0 100644
--- a/docs/irtt-client.md
+++ b/docs/irtt-client.md
@@ -115,7 +115,8 @@ A fixed local port may cause binding conflicts when multiple target sockets are
 
 ## DYNAMIC TARGET SETS FROM STANDARD INPUT
 
-`--targets-stdin` is available only with continuous mode (`--duration 0`).
+`--targets-stdin` is available only with continuous mode (`--continuous` or
+`--duration 0`).
 Positional targets remain an initial desired set, and may be omitted to
 start empty:
 
@@ -208,12 +209,14 @@ Omitting the option preserves the library default schedule.
 `--duration` controls the run length:
 
 - Any nonzero duration (default `10s`) is a **finite** run: probing stops
-  after that long, and a final summary is always printed for table output
-  (see [OUTPUT](#output) below — CSV, TSV, and JSON Lines never print one).
-- `--duration 0` is **continuous**: the client runs until interrupted
-  (`Ctrl-C`). For table output, a final summary is printed only if the run
-  was interrupted. Stdin-controlled mode is the exception: EOF also requests
-  a graceful stop and prints the final retained summary. An uninterrupted
+  after that long, and a final summary is printed for table output when
+  statistics are enabled (see [OUTPUT](#output) below — CSV, TSV, and
+  JSON Lines never print one).
+- `--continuous` (or `--duration 0`) is **continuous**: the client runs until
+  interrupted (`Ctrl-C`). With statistics enabled and table output, a final
+  summary is printed only if the run was interrupted. Stdin-controlled mode
+  is the exception: EOF also requests a graceful stop and prints the final
+  retained summary. An uninterrupted
   continuous run that exits for another reason does not get one.
 
 See [MEASUREMENTS AND MEMORY](#measurements-and-memory) below for why this
@@ -249,6 +252,8 @@ Example for continuous, multi-target CSV output without statistics:
 
 ```sh
 irtt-client --continuous --reconnect --no-stats \
+  --reconnect-timeout 10s \
+  --reconnect-initial-delay 250ms --reconnect-max-delay 10s \
   --format csv --header never \
   --columns client_receive_wall_ns,target,seq,rd_us,sd_us \
   reflector1 reflector2
@@ -344,6 +349,10 @@ processing time is derived from the same server receive endpoint.
 
 ## MEASUREMENTS AND MEMORY
 
+The statistics retention below applies only when the `stats` feature is compiled
+in and `--no-stats` is absent. Disabling collection keeps the client
+classification bounds but allocates no statistics collectors.
+
 Finite and continuous modes make a different tradeoff between exact
 statistics and bounded memory.
 
@@ -394,9 +403,10 @@ the client exits nonzero — so a process supervisor can restart it; an
 interrupted run (`Ctrl-C`) never reports peer closure as an error, whichever
 mode it was in.
 
-With `--targets-stdin`, peer closure is target-local: other desired targets
-and the stdin controller continue running. Including that target in a later
-complete set starts a fresh generation.
+With `--reconnect`, a desired target reopens automatically after peer closure.
+Without recovery, `--targets-stdin` makes peer closure target-local: other
+desired targets and the stdin controller continue running. Including that
+target in a later complete set starts a fresh generation.
 
 ## EXIT STATUS
 
diff --git a/fuzz/Cargo.lock b/fuzz/Cargo.lock
index 3346de1..5dc15e6 100644
--- a/fuzz/Cargo.lock
+++ b/fuzz/Cargo.lock
@@ -132,7 +132,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-proto"
-version = "0.5.1"
+version = "0.5.2"
 dependencies = [
  "hmac",
  "md-5",
@@ -152,7 +152,7 @@ dependencies = [
 
 [[package]]
 name = "irtt-server"
-version = "0.5.3"
+version = "0.5.4"
 dependencies = [
  "getrandom",
  "irtt-proto",