You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
62aa3bc fix(F-886): restart still reaches OUR backend on a two-identity desktop
b604875 F-884: catch the XPath path's trailing dom.disable; correct F-805's half-fix; three low findings
fa05129 fix: the last spawn attempt no longer asks for a re-selection (F-834 review)
1ba7fc7 Merge origin/main (4bf8d62) into test/e2e-lifecycle-resilience; the workspace declares its own log level
1b53726 docs: place F-882b/c under main's Unreleased sections, not above them
dd4b0f7 Merge remote-tracking branch 'origin/main' into fix/F882b-slow-doc-ledger-race
86f791c docs(F-882b/c): correct what the bounded ledger wait still catches
dc6b672 Merge remote-tracking branch 'origin/main' into fix/F884-concurrent-dom-queries
647fb6c Merge remote-tracking branch 'origin/main' into fix/F834-stage1-master-fallback
ecf7eee Merge commit '287228a' into fix/F886-eviction-kills-sibling-browsers
4bf8d62 Merge pull request #125 from DevinoSolutions/fix/F883-execute-script-awaits
287228a test(lifecycle): the strike implication is per PROXY, and a pending verdict is not a silence
9e1ae0c fix(F-883): a DONE Transaction must not be shielded — it recursed; and finish the MQ-128 ledger
298029e fix(F-882c): wait for a fetch the navigation milestone does not cover
d8911e2 fix: every role the resolver issues now has a retry (F-834 stage 1)
6a0720f test: bound the fleet's spawn wave by the cell, and say what the cell allowed
6bad883 fix(F-884): own the wait loop — the per-tab lock must not span nodriver's bundled poll
d016314 fix: the contention hint names the count, not the mechanism (F-834)
e7d704a fix: a spawn that loses the master profile retries onto a clone (F-834 stage 1)
b182e01 Merge branch 'test/e2e-lifecycle-resilience' (42abada) into F-886
42abada Merge origin/main (3bc4e8e) into test/e2e-lifecycle-resilience; S1 asserts the F-820 implication
b2f0b5a chore(F-886): buy back singleton LOC headroom (1000/1000 -> 996)
2b5a08e fix(F-882b): record the slow fixture document on arrival, not after its delay
ec397c0 docs(F-886): correct the residual — session-scoped liveness is NOT safe yet
3503aa8 Merge branch 'test/e2e-lifecycle-resilience' into fix/F886-eviction-kills-sibling-browsers
cd02052 Merge remote-tracking branch 'origin/main' into fix/F886-eviction-kills-sibling-browsers
6aae00f Merge remote-tracking branch 'origin/main' into fix/F884-concurrent-dom-queries
afcfc6a docs(F-886): the second route into the same kill, and the step-aside in S5's print
b6164c7 wip(F-884): docstring for the wait-outside-the-lock rewrite
c401f00 Merge origin/main (F-885b, 3bc4e8e) into fix/F883-execute-script-awaits; CHANGELOG kept under one Unreleased (main's sections first), fixture_routes keeps the F-882 nav block then the F-883 es block appended at the end
3bc4e8e Merge pull request #124 from DevinoSolutions/fix/F885b-remaining-real-log-writes
6837ec2 fix(F-883/F-788/F-794): shield the CDP reply, not the operation — a cancelled caller no longer kills the connection's listener
7ef9def Merge remote-tracking branch 'origin/main' into test/e2e-coverage-F873-F881
4bcb6c4 test: spawn the fleet's master-taker alone, and record why master needs a reservation
f89e510 Merge remote-tracking branch 'origin/main' into fix/F885b-remaining-real-log-writes
1b3364b Merge pull request #121 from DevinoSolutions/fix/F882-navigate-superseded-loader
d2600b8 test(lifecycle): browsers_alive must not read a zombie as alive (review N1)
760875e Merge remote-tracking branch 'origin/main' into fix/F883-execute-script-awaits
09f20dd Merge remote-tracking branch 'origin/main' into fix/F882-navigate-superseded-loader
0a9366f Merge remote-tracking branch 'origin/main' into test/e2e-coverage-F873-F881
091a82a fix: isolate two more real-log-writing test subprocesses (F-885b)
8eff31d test: tolerate the other end of the same slow kill in the fleet's teardown gate
7c0d3f2 fix(F-886): a backend still serving live browsers is never evicted
95affbb Merge pull request #122 from DevinoSolutions/fix/F885-proxy-death-test-isolation
9ff0a85 chore(F-883): drop the temporary B1 probe log committed by mistake
529cec0 fix(F-883): shield _with_cdp_timeout so a late CDP answer cannot kill the tab; key the retry on the record
ec8c5d1 docs: correct the lifecycle E2E claims to what the nodes assert and measured
b6e67c8 test(lifecycle): apply review blockers B1-B6 + nits; deny-list the harness port pick
cf2f286 test: earn the fleet's kill-timeout tolerance with the product's own tracked-pid record
8b18934 F-882 gate: hold the fake's supersession, update the SOFT timeout golden
345a0db fix(F-884): one DOM.getDocument at a time per tab — concurrent resolutions no longer race to -32000
b497d5a test: make the fleet survive a partial spawn and a crowded close, scope its disk claims
1e6ee25 test: F-885 — isolate proxy/backend-death tests from the live ~/.stealth-mcp record
fa420d6 docs: lifecycle resilience E2E — the invariants it guards and the eviction finding
f2e97d4 test: lifecycle resilience E2E — disconnects and browsers closing, over the real wire
68cf1c4 test(F-883): FakeTab models awaitPromise, so two pins can no longer be green for the defect
68ce29f fix(F-882): review nits — section pointers, explicit abort accounting, a networkidle pin that discriminates