Two Budgets Do Not Mean Two Independent Paths
At 17:30 Monday, Gu Ning placed G2's two estimated labor slots and two conditional units beside the six slots of existing work. The margin ledger proved that the quantities fit, but not whether the arrangements depended on the same person, device, or time entrance.
Separate budgets encourage people to imagine separate paths. In practice, one unavailable inspection bench may prevent both from being adjudicated, and an executor occupied by rework on the earlier job may cause the later one to miss its window. Separate project names do not create independent conditions.
This chapter maps dependencies before calculating correlation coefficients or joint-loss probabilities. Shared structure can be checked from schedules and resource records; chances of failure require different material. Declaring risk doubled from one shared node exceeds the evidence just as much as ignoring it.
E4, C1, and M7 are fictional identifiers in Chengwan's ledger. They do not describe actual staffing, equipment reliability, or material patterns in repair work.
Preserve the Six Existing Tasks Under Their Original Identities
The six existing committed slots link to E1 through E6. The identifiers only return to the schedule; they do not assume the jobs are alike or invent revenue, success rates, or technical content.
E4 is scheduled for Wednesday morning and requires Ye Cheng to check its checklist using inspection bench C1. Its commitment predates V2 and remains existing fulfillment protected by Chapter 5.
If G2 is released, its candidate periods are Tuesday afternoon and Wednesday afternoon. The second includes V2's formal inspection and 14:00–16:00 receipt. These times are not reserved; they appear in the exposure map solely to inspect dependencies.
Draw Actions Before Drawing Nodes
Gu Ning draws E4 and G2 as action nodes rather than listing “two projects.” E4 includes Wednesday-morning handling, C1 inspection, and record handoff. G2 includes Tuesday formal preparation, Wednesday final inspection, and receipt.
Each action connects to required conditions: executor, equipment, material entrance, time window, version file, and communication responsibility. A condition serving several actions appears once with edges to each.
This prevents a repeated resource from becoming two. Direction is preserved as well: an action depends on C1, not the reverse. An E4 delay occupying C1 may affect G2 through an ordered connection.
The First Shared Exposure Is Ye Cheng's Time
Ye Cheng executes both E4 on Wednesday morning and G2's formal stage. They appear sequential rather than simultaneous and seem to fit.
But if E4 extends into the afternoon or needs unexpected review, G2's second slot is affected. A retained slot at another time cannot make the same person appear twice on Wednesday afternoon.
What is known is shared personnel and temporal adjacency, not that E4 will be late. The map marks Ye Cheng as a common personnel node, “available, with no confirmed substitute executor.”
One Margin Slot Does Not Create a Substitute Person
Chapter 9 required one slot to remain outside V2 after G2 release. It supplies schedule margin, not a replacement for Ye Cheng. If an anomaly occurs in a Wednesday window that only he can serve, another slot may be too late or unusable by someone else.
Quantities can substitute only when substitutability holds. Treating the retained slot as recovery requires who uses it when, whether they can perform the same task, and whether the client accepts the change. No such material currently exists.
The map therefore separates “one remaining slot” from “Ye Cheng available.” The former retains value without assuming a human-substitution function not established.
The Second Shared Exposure Is Inspection Bench C1
Both E4 and G2 need C1 for Wednesday inspection. The station has no other bench confirmed under both checklists, so C1 has edges to both actions.
Extra review after E4 would remove G2's planned entrance; if E4 remains incomplete, the actions compete for the same time. Separate financial budgets do not separate their equipment condition.
The case does not state that C1 is broken or provide failure history. The node identifies a condition to check; one edge cannot generate a probability.
A Backup Name Is Not a Backup Path
Tang Ke remembered an old inspection bench, C2, in storage. Ye Cheng found no confirmation that it supported the procedures on the two current checklists and no record of its availability this week.
C2 is therefore a candidate recovery resource and cannot be deducted from C1 exposure. Becoming a backup requires at least functional scope, current state, switching time, and an executor able to use it within the window.
Naming an object a backup is easy; proving it can take over when the main path fails is another action. The map retains C1 as the sole confirmed entrance and lists C2's unresolved checks.
The Third Shared Exposure Is Material Entrance M7
Ye Cheng reviews the issue plan and finds that remaining E4 work and candidate G2 purchasing both point to registered batch entrance M7. G1's local sample also used one portion from this entrance and passed S-3.
The G1 pass offers limited support but does not prove every M7 quantity, checklist item, and later time identical. An incorrect entrance identifier, insufficient quantity, or changed condition might affect E4 and G2 together.
No M7 anomaly has been observed. The map records shared source and quantities, identities, and applicability to check; “same batch” does not become “will fail together.”
One Batch Does Not Automatically Mean Complete Homogeneity
Material registered under M7 may contain traceable internal locations, and the portion used by G1 may differ from later portions. Without item-level records, neither complete homogeneity nor independence can be assumed.
Beneath the batch node, the map preserves G1's draw identifier, E4's planned quantity, and G2's planned quantity. Evidence of separate sources or handling can later split the edges; until then, the shared entrance remains.
Granularity serves action. Infinite subdivision manufactures false independence, while indiscriminate combination hides real separation. Gu Ning subdivides only to boundaries the current record can retrieve.
The Fourth Shared Exposure Is Wednesday's Boundary
E4 completes Wednesday morning. G2 inspects Wednesday afternoon and receives from 14:00 to 16:00. Their intraday sequence means any extension of the earlier stage reduces recovery time for the latter.
The common condition is not abstract “time,” but adjacent windows on the same day. Moving E4 to another day may reduce adjacency while affecting other commitments; moving G2 later needs a client-approved version.
Total slot counts cannot reveal temporal exposure. One slot may remain in the week while the fixed 14:00 Wednesday boundary has no movable room. Total and position must both be preserved.
The Fifth Shared Exposure Is the Checklist and Record Entrance
E4 and V2 are different tasks but both depend on the station's version repository for checklist identifiers and inspection records. Tang Ke preserves external confirmations, and Gu Ning adjudicates from those records.
An indexing error may give executors the wrong procedure. A delayed handoff from Tang Ke can affect both record chains. Such sharing may not stop physical work but can make results impossible to adjudicate under the original rules.
Information entrances are resources. Copying a file twice does not create independent sources; if both derive from one erroneous index, both copies preserve the error.
Common Cause Differs from Resource Competition
C1 becoming unavailable would cut both E4 and G2 inspection entrances: a candidate common cause. E4 and G2 seeking the same functioning C1 at the same time is resource competition. Both can obstruct two paths, but require different responses.
A common cause may need replacement equipment or a changed procedure; competition can be addressed through scheduling, priority, or separation in time. Calling both “correlation” does not tell the actor whether to change a node or its order.
The principal visible relations are competition from sequential scheduling and a common entrance in one device. Actual failure still awaits node state and execution results.
Cascading Failure Has a Direction
Suppose E4 finds Wednesday morning that it needs more review, occupying Ye Cheng and C1 into the afternoon. E4 changes first and then compresses G2's window: a cascade from earlier to later work.
If C1 is unavailable at the start of Wednesday, one node affects both E4 and G2, a structure closer to common cause. The scenarios may share an adverse endpoint but have different origins and recovery points.
The map uses arrows for sequence instead of calling every simultaneous problem the same event. When results arrive, Gu Ning can identify which connection should update.
Shared Nodes Can Also Create Coordination Gains
One executor who understands both checklists may reduce handoffs; one bench can standardize records; one material entrance can simplify checking. Sharing is not purely harmful.
Coordination gain and concentration of failure coexist. Replacing every shared person, device, or material adds switching costs and unknowns. Choices compare both sides rather than dismantling every connection.
This chapter exposes structure. Whether to retain or change a node depends on a concrete substitute and cannot be inferred from the number of lines.
Project Diversification and Cause Diversification Are Different
The station has six existing jobs, E1 through E6, plus V2. Yet if many depend on Ye Cheng, C1, or M7, more projects do not create an equal number of recovery paths.
True cause diversification requires another path to retain different and usable critical conditions when one fails. Several task identifiers attached to one resource do not accomplish it.
Conversely, sharing a noncritical condition need not bind every result. The map asks whether a node is necessary, replaceable, and how far its failure propagates.
Dependency Is Not Observed Correlation
The map shows potential mechanistic links: one person cannot act twice at once, one bench cannot be used simultaneously, and one indexing error may propagate. It supplies no cross-batch result data.
Statistical correlation describes variables changing together under defined data and calculation. This chapter has no sample and assigns neither 0.8 to Ye Cheng nor “high correlation” from four edges.
A mechanism map can guide collection; statistical records can test whether joint change occurs. They constrain but do not impersonate each other. Chapter 11 uses independent demonstrations for joint loss; mainline probabilities remain unknown here.
No Edge Does Not Establish Independence
The map contains only dependencies checked so far. No line between actions may mean separate conditions or an undiscovered relation in materials, substitution, or time.
Gu Ning marks unchecked fields unknown instead of reading blank as no dependency. Only after resources, sources, and times are adequately checked can the current use say “no common node found,” still within that inspection scope.
This keeps visual simplicity from creating confidence. Visualization compresses relations and can hide unentered parts. Field sources, not an uncluttered diagram, establish completeness.
Merging and Splitting Nodes Need Grounds
Draw all materials as one “supply” node and every task appears jointly exposed. Draw each M7 portion as wholly independent and the actual batch entrance vanishes. Granularity changes the map's conclusion.
Gu Ning uses retrievable management boundaries: M7 is a common registered entrance with individual quantity identifiers below; C1 is an actual shared device; Wednesday retains ordered morning and afternoon windows. New evidence can split differences or merge a common root.
Finer is not automatically more scientific. A checkable division is more useful than many conjectural nodes. Every split and merge retains its reason so the past structure cannot be redrawn merely to fit outcomes.
Mitigation Actions Also Create New Connections
Validating C2 may require Ye Cheng's learning, Tuesday inspection time, or the same version index. Another equipment node does not guarantee a diversified recovery chain.
Another material entrance may reduce M7 concentration while adding switching procedures and checklist applicability questions. Mitigating one exposure changes consequences elsewhere; a line cannot simply be erased.
Each mitigation therefore enters as a candidate action: what it needs, when it completes, and what its failure affects. Only actual validation and authorization update the main path. Risk treatment also accepts action discipline.
One Pass Cannot Guarantee Shared Materials
G1's M7 portion passed S-3, adding material beyond pure assumption. It supports a local relation under the same identifier and procedure, not every E4 and G2 use.
Calling this “M7 validated” would turn the node into false protection. A later deviation in another item might then be assigned automatically to execution.
The map marks the scope of G1 evidence and the quantities, items, and times not covered. Favorable material enters without losing its boundary.
Ask Four Questions About Every Edge
Each dependency edge must answer: is the node necessary; when is it used; is an existing substitute available in the same adverse scenario; and how does node change propagate to the result?
Ye Cheng is the current main executor on Tuesday and Wednesday, with no confirmed substitute; his unavailability pauses the corresponding action. C1 is required for Wednesday inspection; C2 is unvalidated; C1's unavailability prevents checklist adjudication.
M7 quantities and checklist correspondence await review. Time comes from the client window and E4 sequence. Original version records are retrievable, but their handoff remains concentrated in Tang Ke. The four questions give each line practical meaning.
Recovery Paths Also Need a Common-Root Check
If C1 is unavailable, asking Ye Cheng to work faster does not restore it. If Ye Cheng is unavailable, moving a task to the same time does not help. A recovery measure must act on the failed node.
Two backups both requiring C1 are not two paths for C1 failure. Even if C2 is validated, personnel remain concentrated if only Ye Cheng can use it. A backup must be checked along the entire critical chain.
G2 currently has no fully validated alternative path. Survival does not require it to continue. If a critical node changes, stopping G2, protecting E4, and treating the balance under the terms remain an actual exit.
Set Priority Before Conflict Occurs
Chapter 5 placed existing fulfillment among hard constraints. If E4 and G2 compete for Ye Cheng or C1 Wednesday, E4 takes priority; G2 pauses and reopens the receipt window.
Recording this before conflict prevents a received payment or approaching delivery from promoting the new task. It also makes G2's adverse scenario describable: V2 may miss its window, but E4 is not displaced to recover it.
Priority creates no resources. It decides what to protect when scarcity occurs and places the other path's stopping consequence in the plan.
Can Rescheduling Reduce the Connection?
Gu Ning tries placing G2's first slot Tuesday afternoon and keeping its second Wednesday afternoon. This avoids simultaneous use with E4 in the morning but preserves adjacency and shared C1.
Moving G2's second slot to Tuesday would create new storage and handoff conditions before Wednesday receipt, and the current checklist does not say whether that is valid. A prettier map cannot authorize a changed process.
Moving E4 alters an earlier commitment and requires agreement from its responsible people and object. No such material exists, so the exposure map preserves the candidate sequence rather than assuming rescheduling occurred.
Node Checks Can Precede Resource Release
Gu Ning adds three G2 entry checks. Before Tuesday's start, confirm M7's identifier, quantity, and relation to G1. G2's Wednesday segment continues only after E4 releases Ye Cheng and C1. Tang Ke checks the version index before each action.
These checks reduce some misuse without eliminating the Ye Cheng single point or C1 failure. They consume limited time and enter execution records when performed; they are not costless protection.
G2 remains unreleased. The additions revise its gate and state what starting and continuing depend on.
Checking Also Consumes Limited Attention
Reviewing M7, C1, and the version index needs people and time. If every imaginable common cause is listed, Tuesday's available slot may be consumed by analysis itself.
Gu Ning prioritizes direct controls over G2 entry and Wednesday continuation by necessity, substitute status, and urgency. Remote hypotheses without an identifiable current path do not expand indefinitely in this round.
This does not assume unchecked parts safe. It states the inspection scope for a limited decision period. A later omission becomes a model boundary in review; “caution” cannot hide analysis cost.
Exposure Severity Is Not Ranked by Edge Count
A node with four noncritical edges may matter less than one edge on an irreplaceable path. Edge count can signal concentration but is neither loss magnitude nor probability.
Gu Ning separately records necessity, substitute state, urgency, and affected protected object. Ye Cheng and C1 are necessary Wednesday and have no confirmed substitute, so they take priority. Version originals are retrievable and recovery clearer. M7 still needs quantity and applicability checks.
This ranking is a working judgment in the case, not a precise risk score. Evidence that C2 or other personnel are usable must update node state.
Jointly Bad Results Do Not Prove a Common Cause
If both E4 and G2 later fail, C1 may be a common cause; they may also have distinct causes, or one may contain execution deviation while the other has incompatible materials. A shared endpoint begins investigation rather than answering it.
Likewise, both passing does not prove independence. Shared structure remains when the common node happens not to fail. Whether it persists in a later batch depends on changed resources and processes.
Attribution must return to process records, node states, and time order. The map helps avoid prior omission but cannot announce a posterior cause.
Multiple Paths Must Check Their Common Convergence Points
RC calls for heterogeneous information sources and multiple realization paths in the spatial dimension, not to inflate nominal quantity but to prevent one observational frame from locking every possibility. Paths all depending on Ye Cheng, C1, M7, and one window still run through a common convergence channel at critical points.
The exposure map therefore tests structural “path diversity.” It distinguishes margin from heterogeneous relations from several names for one dependency. Only by seeing common and independent nodes can multiple paths preserve real room for reconvergence when one fails.
The D17 Shared-Exposure Map
At 18:00 Monday, Gu Ning saves the main map. Action nodes are E4, G2's Tuesday segment, and G2's Wednesday segment. Shared nodes are Ye Cheng, C1, M7 entrance, Wednesday's adjacent windows, and the version-record entrance.
Ye Cheng connects to E4 and both G2 segments; C1 to E4 inspection and G2 Wednesday inspection; M7 to E4's material plan, G1's used identifier, and G2's candidate quantity; the Wednesday window to E4 completion and V2 receipt; the version entrance to execution and adjudication.
Each edge names required time, substitute state, responsible person, and propagated result. Sharing comes from the current schedule and records; node-failure probabilities remain unknown. C2 stays a candidate recovery resource rather than an effective backup.
How Exposure Review Changes G2
The budget had required two slots, two units, and one retained slot. The map adds sequence: G2's Wednesday segment waits for E4 to release Ye Cheng and C1. If unreleased at 11:00, G2 does not take the margin slot automatically; it pauses and reopens the client's window.
Material conditions tighten as well: before G2's first segment, check the exact M7 identifier, quantity, and transfer from G1 evidence. Tang Ke must check the shared original-record entrance rather than a relayed copy.
These changes do not make G2 pass. Its state moves from “shared exposure pending” to “exposure structure identified; node states and joint consequences pending,” still without reservation, purchase, or production.
Handing Off to the Joint-Outcome Table at 18:20 Monday
At 18:20, the station completes the exposure map. There is no new external reply; the first-payment balance and fourteen-unit snapshot remain; three slots remain uncommitted; and E4 has neither moved nor been canceled.
The main finding is not that every path will fail together, but that independence has not been established. Separate budgets remain connected by Ye Cheng, C1, M7, time, and records. Finding the connections locates recovery and observation without inventing a failure probability.
The next chapter uses explicitly labeled independent demonstrations to compare different joint structures under equal marginal loss rates, then returns the method to D17. Project count, average loss, and common nodes answer different questions. When joint material is inadequate, the station retains ranges and scenarios instead of calling nominal diversification verified protection.