cf-docs-rfc-proposal

Intent: troubleshooting_slow_requests

Pattern: result / explanation / action

All six tables have the same three columns, one table per experiment. They share one set of hints:

Level Hint Value From source
Table pattern result-explanation-action same headers in all six tables
Table header row yes <thead>
Column 1 role key “Result”: the outcome each row is about
Column 1 width about 25% width="25%" on a column 1 cell in Tables 2–6
Column 1 wrap normal results are sentences, not identifiers
Column 2 role prose “Explanation”
Column 2 width about 25% width="25%" on a column 2 cell in Tables 5 and 6
Column 3 role prose “Action”
Column 3 width the rest (about 50%)  

Width variants. The source sets widths inconsistently: Table 1 sets none, Tables 2–4 set column 1 only, and Tables 5 and 6 set columns 1 and 2. The widths sit on one body cell and apply to the whole column only because browsers size a column from its widest request. Three variants are rendered and compared before choosing:

Variant Widths Where
A — as written each table’s own widths passthrough
B — uniform about 25% / 25% / 50% in all six tables extension
C — roles only no width hints; the key and prose defaults decide extension

Decided after comparing renderings: variant C. The width rows in the hints table above no longer apply; the roles decide. Passthrough still carries each table’s widths as written (it applies cleanup only).

Cleanup needed (all tables): none for well-formedness. Body rows sit directly in <table> with no <tbody>; browsers insert one. Adding <tbody> changes no content.

Table 1 — Experiment 1: total round-trip app requests

Source: lines 69–100. Four rows. No widths in the source.

Cleanup needed: none.

Spot checks

Table 2 — Experiment 2: request time in access logs

Source: lines 149–183. Four rows. Source width: column 1, row 1 and 2.

Cleanup needed: none.

Spot checks

Table 3 — Experiment 3: duplicate latency on another endpoint

Source: lines 210–229. Two rows. Source width: column 1, row 1.

Cleanup needed: none.

Spot checks

Table 4 — Experiment 4: remove the load balancer

Source: lines 275–296. Two rows. Source width: column 1, row 1.

Cleanup needed: none.

Spot checks

Table 5 — Experiment 5: remove Gorouter

Source: lines 330–350. Two rows. Source widths: columns 1 and 2, row 1.

Cleanup needed: none.

Spot checks

Table 6 — Experiment 6: network between router and app

Source: lines 395–415. Two rows. Source widths: columns 1 and 2, row 1.

Cleanup needed: none.

Spot checks

Page-level checks