fix(udps): stop packets being dated from an earlier time base

Reported as samples sporadically carrying a previous packet's timestamp:
holes on one side of the stream and collisions on the other, in both the
Go and the MARTe2 receiver. That it appeared in both is what located it
-- the shared cause is upstream of either client. Four independent
defects, all of which end in a packet's values being placed at a time
that is not theirs.

Reassembly slot exhaustion (the "Reassembly slots full; evicting oldest"
flood). Chunk size was learnt only from fragment 0, so an out-of-order
burst destroyed a packet whose bytes had all arrived and left the slot
occupied until the 2 s GC. Slots were keyed on the counter alone, but
DATA and CONFIG number independently, so equal counters merged the two
streams. The 32-byte received-mask covered 256 of the 512 fragments the
client accepts, so a duplicate above 255 was counted as new and the
packet was delivered with a hole of stale bytes in it. And one datagram
was read per Execute(), which cannot drain a fast producer. Fixed with a
pendingTail deferral, (counter, type) keying, a 64-byte mask, a
256-datagram drain, counter-age slot reclamation, and a 1 Hz aggregated
warning in place of the per-eviction flood.

UDPStreamer dropping whole Accumulate batches. EventSem::ResetWait is
Reset-then-Wait, so a Post() landing while the sender thread was inside
ServiceClients()/SendData() was destroyed by the next Reset. The batch
was then skipped with dataReady false, readyFill was never cleared, and
the following flush overwrote it: an entire run of RT cycles never
reached the wire. The record of pending work now lives in the buffers
rather than in the semaphore edge, which also removes up to
UDPS_DATA_WAIT_MS of latency; genuine backpressure overwrites are
counted and reported. Against the unfixed code the new test sees
2999/3000 batches never consumed.

Period inflation after loss. Accumulated scalars carry no SamplingRate,
so the receiver derives dt from the sender-clock gap -- but dividing it
by the previous packet's sample count is only right while nothing is
lost. One loss doubles the reported period, which spreads a batch a full
batch past its own end and into the range the next packet claims. That
is the hole and the collision, exactly. Inferring the cycle count from
the estimate's own period is not a way out: it has a stable fixed point
wherever gap/dt is an integer, so a real rate change locks it at the old
one for good (AccumDtGTest.FollowsSustainedRateChange).

The packet counter removes the ambiguity, so all three receivers now
order on it: a DATA packet that does not advance the counter is dropped
rather than delivered, because its values are older than data already
handed over. Ordering is on the signed difference so it survives the
uint32 wrap, and the sequence resets on reconnect, where the producer's
counter restarts independently of ours. The loss count that falls out of
the same delta feeds the period estimate as cycles = prevN * (1 + lost),
which reduces exactly to gap/prevN when nothing is lost and therefore
still tracks a genuine rate change. UDPSClient::AcceptDataCounter (C++),
udpsprotocol.SequenceGate (Go), decode_data (C).

The C client's existing gap counter was wrap-unsafe and let a stale
packet rewind last_counter, which made every subsequent gap wrong; it
uses the same code now. Docs/Protocol.md gains an Ordering DATA section
stating the requirement for any receiver, including ones outside this
repository.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Docs/Protocol.md

@@ -171,6 +171,39 @@ the client to reassemble them in any order.
 
 ---
 
+## Ordering DATA (required of every receiver)
+
+DATA carries its own `counter` sequence, incremented once per sent packet
+(CONFIG is numbered independently). Reassembly completes in arrival order, not
+counter order, so a packet reordered or duplicated on the wire surfaces after a
+newer one has already been consumed. Its values are well-formed but carry an
+older time base: accepting it writes them over samples the consumer already
+holds and leaves the span they should have filled empty — a collision on one
+side and a hole on the other.
+
+A receiver must therefore drop any DATA packet that does not advance the
+counter, and must order it by the *signed* difference:
+
+```c
+int32_t delta = (int32_t)(counter - lastCounter);   /* survives the uint32 wrap */
+if (delta <= 0) { /* stale or duplicate: drop */ }
+lost = (uint32_t)delta - 1u;                        /* packets missing before this one */
+```
+
+Comparing the values directly would call the first packet after the wrap stale
+and reject the stream from then on.
+
+`lost` matters beyond diagnostics. A consumer that spaces batched samples from
+the elapsed time since the previous packet must divide that gap by `lost + 1`
+batches; dividing by one batch reports a period too long by exactly that factor
+and walks the samples past their own end into the next packet's range. Reset
+the sequence on (re)connect: the producer's counter restarts independently.
+
+Implemented in `UDPSClient::AcceptDataCounter` (C++),
+`udpsprotocol.SequenceGate` (Go) and `decode_data` (C).
+
+---
+
 ## Minimal Python Client Example
 
 ```python

Docs/UDPS-C-Client.md

@@ -175,6 +175,7 @@ replayed traffic can be decoded without a client.
 ```c
 typedef struct {
     uint32_t counter;      /* gaps in this sequence are lost datagrams */
+    uint32_t lost;         /* DATA packets missing immediately before this one */
     uint64_t hrt;          /* producer's high-resolution timer at send */
     double   recv_time;    /* CLOCK_REALTIME seconds at arrival */
     uint8_t  publish_mode;
@@ -194,6 +195,13 @@ scalar signal in Accumulate mode, where the producer batches several RT cycles i
 and `count == num_samples` — one value per cycle. Arrays are not batched: they appear once and
 apply to the whole packet. `udps_frame_value(f, sig, sample, elem)` applies that rule for you.
 
+**Ordering.** Frames reach `on_data` in counter order: a DATA packet that does not advance the
+counter — reordered or duplicated on the wire — is dropped rather than delivered, because its
+values carry a time base older than data you already have, and placing them would overwrite live
+samples while leaving their own span empty. `lost` reports how many packets went missing just
+before the frame. If you space samples yourself from the elapsed time since the previous frame,
+divide by `lost + 1` batches, not one: the gap covers the missing packets' cycles too.
+
 **Timestamps.** The protocol does not put a timestamp on every element; how to date them depends
 on the signal's `time_mode` (see [Protocol.md](Protocol.md#time-mode-codes)):
 
@@ -223,6 +231,7 @@ udps_client_stats(cli, &s);
 | `frames_delivered` | DATA packets decoded and passed to `on_data`. |
 | `config_updates` | CONFIG packets applied. |
 | `counter_gaps` | Missing packet counters — datagrams lost on the wire or in the kernel. |
+| `stale_packets` | DATA packets dropped for not advancing the counter: reordered or duplicated on the wire. |
 | `fragments_dropped` | Duplicate, stale or unplaceable fragments; a non-zero value with `counter_gaps` means fragmented updates are arriving incomplete. |
 | `reconnects` | Sessions re-established after a silence timeout. |