What a plan check is
An applicant submits drawings to the authority having jurisdiction. Reviewers check them against the adopted codes, produce a correction letter listing every deficiency, and the applicant responds with revised sheets and a written response to each comment. The reviewer rechecks, issues fewer comments, and eventually approves.
The IBC's administrative chapter describes this in a paragraph. In practice it takes months, and the months are not evenly distributed.
Review is parallel, approval is serial
A commercial submittal does not go to one reviewer. It goes to building, structural, mechanical, plumbing, electrical, fire and planning, and often to public works, health and accessibility as well, each of which reads the same set looking for a different thing.
Those reviews run in parallel, which sounds efficient until you notice the permit issues only when the last one clears. Seven reviewers who each take three weeks produce a three-week review, unless one of them takes seven, in which case the whole submittal takes seven and the other six sat idle for a month.
Approval is governed by the slowest reviewer, and the slowest reviewer is rarely the same discipline twice.
Each resubmittal restarts the clock
Here is the mechanic that surprises applicants most. In many jurisdictions a resubmittal enters the queue as a new submission rather than resuming where it left off.
That means a response to two minor plumbing comments can cost another full review cycle across every discipline, because the file is reassigned and the reviewers who already approved may recheck against the revised sheets. Four rounds of trivial corrections can consume more calendar time than the original review did.
The practical consequence: consolidate. A single complete resubmittal that answers every comment from every discipline beats three fast partial responses, and applicants who send corrections as they finish them are lengthening their own permit.
What the comments are actually about
Correction letters look like code citations and read like a list of unrelated problems. Sort a few hundred of them and the same categories keep appearing.
A large share are completeness comments rather than compliance comments: a missing sheet, an unreferenced detail, a schedule that does not match the plan, a specification section cited on the drawings but absent from the submittal. The design is fine. The package is not, and the reviewer cannot approve what they cannot find.
Another large share are coordination comments, where two disciplines describe the same condition differently. The architectural reflected ceiling plan shows a diffuser the mechanical drawings do not, or the structural drawings carry a beam depth the ceiling height cannot accommodate. A reviewer catches it because reading four disciplines against each other is the job.
Genuine code disagreements, where the designer read a section one way and the reviewer read it another, are the smallest category and the only one that needs a professional conversation. Most of a correction letter is bookkeeping.
What an applicant controls
Package completeness is the biggest lever and the least glamorous. Running an internal check against the jurisdiction's own submittal checklist before filing removes a category of comment entirely, and most jurisdictions publish that checklist.
Cross-discipline coordination is the second lever. Clash detection catches the geometric conflicts; the ones that survive are usually text against geometry, such as a door schedule that disagrees with a floor plan, and those need someone reading rather than a model checking.
The response letter is the third and it is undervalued. A reviewer rechecking a resubmittal is looking for their own comment, the applicant's response, and the sheet where the change lives. A response that says "revised" with no sheet reference makes the reviewer hunt, and a reviewer hunting through 200 sheets is a reviewer who takes another week.
Where the delay really accumulates
Add it up and the pattern is consistent. A first review takes its published time. The correction letter arrives with sixty comments, most of them completeness and coordination items. The applicant answers forty, defers twenty pending a consultant, and resubmits partially. That partial resubmittal restarts a full cycle. The second letter has eighteen comments, twelve of which are the deferred ones plus new comments generated by the revised sheets.
Four rounds later the permit issues, and the project team records the delay as jurisdiction slowness. The jurisdiction spent perhaps six weeks reading. The other four months were queue time created by the resubmittal pattern.
An applicant who consolidates responses, references sheets precisely, and checks their own package against the published checklist can often collapse four rounds into two. That is a schedule change measured in months, achieved with no code argument and no expediter.
The unresolved part
None of this fixes the underlying asymmetry, which is that a reviewer sees the package once and the design team has lived in it for a year. The reviewer's questions are legitimate precisely because they are the questions a stranger asks.
The best applicants write for that stranger. They do not assume the drawing set explains itself, because to the only reader who matters, it does not.