threading depth analysis

summary metrics

metric	value
total threads	4,656
total edges (connections)	2,562
root threads (spawn origins)	208
max chain depth	72
avg chain depth	4.58
orphan count	2,911
orphan rate	62.5%

edge types

type	count	% of edges
continuations (“Continuing work from…“)	1,824	71.2%
handoffs (“Created handoff thread…“)	203	7.9%
read_thread references	535	20.9%

depth distribution

most threads are shallow, but a fat tail extends to depth 72:

depth 0:  3119 threads (67%)  ████████████████████████████████████████
depth 1:   308 threads (7%)   ████
depth 2:   297 threads (6%)   ████
depth 3:   245 threads (5%)   ███
depth 4:    94 threads (2%)   █
depth 5:    81 threads (2%)   █
depth 6-10: ~40/level
depth 11-30: ~12/level
depth 31-72: 1-8/level (single long chains)

top spawners (most children)

threads that spawned the most child threads:

rank	thread	children
1	T-019b37e4-86b2-7079-a4a5-4156a30fda88	106
2	T-019b3650-66cf-74ab-bf0b-eab5e947ae70	79
3	T-019b9a3d-f7bc-74b5-b360-4fa4d12e1a8e	59
4	T-019b99e7-d4c7-726e-9585-55db7fc4add8	27
5	T-019b99e2-192e-7545-be0c-4b7ec9df12c5	23
6	T-019b37a2-8003-7761-8062-8099eaae05b5	19
7	T-019b385a-d8b2-71c3-886d-fa6f1f39c77b	14
8	T-019b523c-7743-7777-9e95-a42f6eac175a	14
9	T-019b9a9b-6430-71dd-b927-0458a46702f6	13
10	T-019b6ba2-1a1c-70dd-ba4e-d4c5b08cb04d	12

observation: top spawner has 106 children - this is a coordinator thread running parallel analysis waves.

longest chains

the deepest chains all share a common root: T-019b8564-338c-736b-9905-0ad763d2216e

this represents a marathon migration task that handoffed 72 times - likely a large icon migration or similar repetitive batch work.

sample chain (depth 72)

T-019b8564 (root)
  └─ T-019b869a (depth 1)
       └─ T-019b88c5 (depth 2)
            └─ T-019b88d0 (depth 3)
                 └─ ... (continues 68 more levels)
                      └─ T-019b9964 (depth 72)

productive spawn patterns

pattern 1: parallel fan-out (wide)

single coordinator spawns 50-100+ agents
each agent completes independent task
coordinator collects results
example: T-019b37e4 spawning 106 analysis agents

pattern 2: sequential handoff (deep)

task too large for single context
each thread completes chunk, hands off to next
maintains continuity via “Continuing work from” references
example: the depth-72 migration chain

pattern 3: hybrid (wide + shallow)

coordinator spawns ~20 agents
each agent may spawn 1-3 sub-agents
typical depth: 2-4
most common pattern in codebase

orphan analysis

62.5% orphan rate seems high but is expected:

standalone tasks: user starts thread, completes in one context
exploratory threads: quick questions, no follow-up
failed spawns: agent crashed/cancelled before connecting

the 37.5% connected threads represent meaningful multi-step work.

insights

depth matters for complexity: depth 72 is extreme - suggests task decomposition could be improved. ideal chains should be <10 deep.
wide > deep for parallelism: the 106-child spawner is more efficient than the 72-deep chain. parallel work completes faster, easier to recover from failures.
read_thread underutilized: only 535 read_thread calls across 4,656 threads. agents could benefit from more cross-thread context sharing.
handoffs are expensive: only 203 handoffs vs 1,824 continuations. handoffs involve more ceremony (creating new thread, passing context). continuations are lighter.

recommendations

break deep chains earlier - spawn parallel workers instead of sequential handoffs after depth ~5
use read_thread more to reference prior work without creating parent-child edges
for batch migrations: spawn parallel batches, use coordinator to aggregate
monitor orphan rate as health metric - sudden spike may indicate spawn failures